Monday Aug 3
FUNDINGNVIDIA

Nvidia Pours Billions Into Sutskever's SSI

3AUG
$100B BETNVIDIASSI

Ilya Sutskever's secretive AI lab just broke two years of silence. Nvidia signed a multi-billion dollar deal for Vera Rubin chip access. SSI still has zero products, yet investors keep piling in.

Sutskever frames the money as scaling proven research, not chasing a shipping deadline. Compute jumps roughly tenfold within twelve months on Nvidia's newest GPU generation.

SSI has already raised seven billion dollars total. It now carries a valuation near thirty two billion dollars, with no shipped product. Nvidia was already an investor before this compute agreement, deepening its bet on Sutskever's team.

Critics call the research "worthy of scaling" a promise rather than proof. SSI has shipped nothing public in two years of operation. Nvidia is betting raw compute now outweighs an actual track record.

full brief & sources

⚡ Why this matters

  • Sutskever's departure from OpenAI in 2024 was one of the most dramatic exits in AI history.
  • SSI has raised $7 billion without releasing a single product or paper.
  • Nvidia backing a pure-research lab signals compute is now the scarce resource, not ideas.

🔍 What happened

  • Nvidia and SSI announced a long-term strategic partnership on July 27.
  • The deal includes an undisclosed investment stretching into multiple billions.
  • SSI gets access to Nvidia's next-gen Vera Rubin GPU platform.
  • Compute capacity increases by "an order of magnitude" over 12 months.
  • SSI has raised $7 billion total and carries a $32 billion valuation.
  • Nvidia was already an investor before this new compute deal.

💬 Smart takes

  • Ilya Sutskever, Co-founder, SSI: "We have research that is worthy of scaling up, and having access to a big Nvidia computer will let us do so."
  • Nvidia: the partnership will "accelerate SSI's next stage of growth after obtaining rare access into the company's closely guarded research."
  • Skeptic: SSI has shipped nothing in two years, so "worthy of scaling" is still a promise, not proof.

🧭 Where this goes

  1. LikelySSI keeps its research under wraps even after the compute boost, true to its "straight shot" philosophy.
  2. LikelyNvidia uses the deal to show it's backing multiple horses in the alignment race, not just OpenAI and Anthropic.
  3. PossibleSSI publishes its first paper or benchmark within the next year, ending the silence.
  4. Wild CardSSI merges with or gets acquired by a bigger lab once its compute runs out.

🥄 The Spoon Take

SSI raised billions on reputation alone, with zero shipped products in two years. Nvidia backing it anyway shows compute has replaced traction as the real signal of AI credibility. That's either a huge bet on Sutskever, or proof the funding market has decoupled from results.

🤔 Pushback

Maybe SSI really is different, and patient capital on alignment research is exactly what the industry needs right now.

·TechCrunch - Nvidia backs Sutskever's SSI
Friday Jul 17
FUNDINGOther

Fireworks Raises $1.5B At $17.5B

17JUL
40T TOKENS$1.5B RAISED$17.5B VALUE

AI infrastructure is turning into serious money, fast. Fireworks just raised $1.5 billion at a $17.5 billion valuation. Daily token volume nearly tripled to 40 trillion, proving real demand, not hype.

Lin Qiao ran PyTorch at Meta before co-founding Fireworks. Her bet: most firms need a fast, cheap model, not a frontier one.

Fireworks now serves 40 trillion tokens a day, up from 15 trillion a year ago. Revenue crossed $1 billion in annualized run rate. Uber, Shopify, and Revolut already run production workloads on it.

NVIDIA is both an investor and infrastructure partner here, not just a cloud vendor. That's a bet that specialized inference beats general-purpose cloud AI at scale.

full brief & sources

⚡ Why this matters

  • AI infrastructure spend is shifting from experiments to production, and Fireworks is where that money lands.
  • A 5x revenue jump and a near-tripling of token volume year-over-year is real usage, not a valuation story alone.

🔍 What happened

  • Fireworks announced a $1.505 billion Series D on July 16, 2026, at a $17.5 billion valuation.
  • The company surpassed $1 billion in annualized revenue run rate, up 5x year-over-year.
  • Daily tokens served nearly tripled, from 15 trillion to more than 40 trillion.
  • Co-founder Lin Qiao previously led Meta's PyTorch team.
  • Customers running production workloads include Uber, Shopify, and Revolut.
  • NVIDIA, Index Ventures, Bessemer, Insight Partners, and Lightspeed are among the investors.

💬 Smart takes

  • Fireworks: positions itself as leading the 'specialized intelligence revolution' - tuned models beating general-purpose ones on cost and speed.
  • Skeptic: $17.5 billion for an inference layer assumes frontier labs won't just undercut them on price once compute gets cheaper.

🧭 Where this goes

  1. LikelyFireworks uses the round to expand compute capacity ahead of a possible IPO push.
  2. Likelymore enterprises shift routine AI workloads from general models to task-tuned ones for cost reasons.
  3. Possiblea frontier lab launches a competing 'cheap tuned model' tier to defend share.
  4. Wild CardFireworks or a rival gets acquired by a hyperscaler within 18 months to lock in the infra layer.

🥄 The Spoon Take

Every AI inference dollar spent isn't going to the flashiest model, it's going to whatever's cheapest that still works. Fireworks just proved that market is worth $17.5 billion and growing 5x a year. The real AI gold rush might be in the plumbing, not the chatbot.

🤔 Pushback

Token volume and revenue are self-reported ahead of a funding round, and neither figure is independently audited yet.

·SiliconANGLE - Fireworks $1.5B round·Fireworks - Series D announcement·Yahoo Finance - Fireworks Series D
Sunday Jul 12
FUNDINGOther

SambaNova Raises $1B, Hits $11B

12JUL
SN50 CHIP$1B RAISE

AI chips got a much bigger war chest. SambaNova raised $1 billion at an $11 billion valuation, led by General Atlantic. JPMorgan Chase signed on to run its chips for secure, on-site AI.

Five months after a $350 million round, SambaNova is already back for more. BlackRock, Intel Capital, and Qatar's sovereign fund all joined this one.

The pitch: banks that can't use the public cloud need AI chips on-site. SambaNova's SN40L and SN50 now handle that inside JPMorgan - a direct swing at Nvidia's enterprise grip.

Investors clearly expect more than one winner in AI chips. Inference, not training, is where the next fight happens.

full brief & sources

⚡ Why this matters

  • Inference chips, the ones that run trained models rather than train them, are becoming the bigger market.
  • JPMorgan Chase choosing SambaNova over Nvidia for on-site inference is a real enterprise vote of confidence.
  • A second $1B+ round in five months shows investors still expect a multi-winner chip market, not just Nvidia.

🔍 What happened

  • SambaNova closed the first tranche of a $1 billion Series F on July 8, 2026.
  • The round values the company at $11 billion, up from a $350 million round in February.
  • General Atlantic led; BlackRock, Intel Capital, Qatar Investment Authority, and Vista Equity Partners also joined.
  • JPMorgan Chase will run SambaNova's SN40L and SN50 chips for secure, on-premises AI inference.
  • SN50 chips are built to handle trillion-parameter models for on-premise deployment.

💬 Smart takes

  • Martin Escobari, General Atlantic co-president: 'SambaNova's platform is differentiated, built for a market where inference has become foundational to enterprise and industry transformation.'
  • Skeptic: Nvidia still owns the software stack most enterprise AI teams already build on - swapping chips is harder than swapping vendors on paper.

🧭 Where this goes

  1. LikelySambaNova announces 2-3 more bank or regulated-industry customers within two quarters.
  2. LikelyNvidia responds with its own on-premises inference push aimed at the same banks.
  3. PossibleSambaNova files for an IPO within 18 months, following Cerebras's public listing.
  4. Wild Carda major cloud provider acquires SambaNova outright to shortcut its own inference chip roadmap.

🥄 The Spoon Take

Every AI infra story from 2024 was about training bigger models. This one is about running them cheaply, on-site, for regulated industries that can't use the public cloud. JPMorgan picking SambaNova over Nvidia for inference is the more interesting fact than the $11 billion number.

🤔 Pushback

Nvidia's software moat is much harder to displace than a single hardware win at one bank suggests.

·TechCrunch - $1B raise at $11B·General Atlantic - Deal announcement·AI Insider - JPMorgan partnership
Sunday Jul 5
FUNDINGOpenAI

OpenAI May Wait On Its IPO

5JUL
2026 PLANOPENAI WAITS

OpenAI filed to go public. Now it might wait. Advisers to CEO Sam Altman reportedly see SpaceX's rocky debut as a bad omen. Anthropic, which filed later, could end up going public first.

SpaceX opened at $150, spiked to $225, and slid back to around $156 within two weeks of its IPO. That swing is reportedly spooking OpenAI's advisers.

OpenAI filed confidentially in June, aiming for a September debut near a $1 trillion valuation. A delay to 2027 would flip that plan.

OpenAI lost about $21 billion last year against $13 billion in revenue. That math gets more public scrutiny the longer the company waits.

full brief & sources

⚡ Why this matters

  • A delayed OpenAI IPO could hand Anthropic the symbolic 'first AI lab public' milestone.
  • It signals real nerves about how public markets price AI spending versus AI profit.
  • Timing here sets the comp for every other AI IPO waiting in line.

🔍 What happened

  • The New York Times reported Altman's advisers are citing SpaceX's volatile debut as a warning sign.
  • OpenAI confidentially filed for an IPO in early June, targeting September 2026.
  • Cerebras, another recent AI IPO, has also stayed volatile since its debut.
  • OpenAI's 2025 operating loss was near $21 billion on about $13 billion in revenue.
  • Anthropic filed its own confidential S-1 on June 1, ahead of OpenAI's filing.

💬 Smart takes

  • Motley Fool analysis: OpenAI is trying to avoid SpaceX's volatility, and will probably not succeed, since large IPOs are volatile almost by nature.
  • Skeptic: Jefferies data shows big IPOs average 26.5% first-week gains but only 3.5% after a year, so timing the market rarely works out as planned.

🧭 Where this goes

  1. PossibleOpenAI pushes its debut into 2027, letting Anthropic go public first.
  2. PossibleOpenAI proceeds on the original September timeline despite the internal debate.
  3. Wild Cardboth labs delay, and a smaller AI company ends up as the first pure-play frontier lab to trade publicly.
  4. Likelywhichever lab goes first sets the valuation multiple every other AI IPO gets measured against.

🥄 The Spoon Take

Being first to file isn't the same as being first to ring the bell. If OpenAI waits, Anthropic inherits a symbolic win it didn't even have to fight for. Markets reward whoever proves the AI-spending story works, not whoever files first.

🤔 Pushback

This is one report citing unnamed advisers. OpenAI could still list in September and this delay chatter fades by next week.

·Motley Fool - OpenAI 2027 IPO report·TheStreet - $1T valuation holdout·Fortune - Anthropic files confidential S-1 first
Saturday Jul 4
FUNDINGOther

Together AI Raises $800M, Hits $8.3B

4JUL
CLOSED$8.3B

Silicon Valley's closed AI startups just got outraised by an open alternative. Together AI, which runs models like DeepSeek, landed $800 million. Saudi Arabia's Aramco Ventures led the round at $8.3 billion.

Founder Vipul Ved Prakash built Together AI to run open models cheaply. The company now books over $1 billion a year.

Open-model usage on the platform tripled in the past year. Nvidia, Salesforce, and SentinelOne all joined this round as investors. The plan: grow infrastructure roughly 50-fold over five years.

A Saudi oil fund leading an AI infrastructure round is new. Watch whether more sovereign money chases the open-source layer next.

full brief & sources

⚡ Why this matters

  • Sovereign capital is now betting on open-source AI infrastructure, not just frontier labs.
  • Signals real revenue in the layer beneath the model race, not just the models themselves.
  • Open-weight inference at scale is looking like a durable business, not a stopgap.

🔍 What happened

  • Together AI raised $800 million in a Series C, closing July 1, 2026.
  • The round set an $8.3 billion post-money valuation, up sharply from its prior round.
  • Aramco Ventures, the venture arm of Saudi Arabia's state oil company, led the round.
  • Nvidia, Vista Equity Partners, General Catalyst, Salesforce Ventures, and SentinelOne's S Ventures also joined.
  • Together AI runs open models like DeepSeek, Nemotron, and Kimi for enterprise customers.
  • The company says it already books over $1 billion a year and plans to scale infrastructure roughly 50x in five years.

💬 Smart takes

  • Together AI: says usage of open-source models on its platform tripled over the past year.
  • Skeptic: one infra vendor's bookings number doesn't prove open models are beating closed ones on quality.

🧭 Where this goes

  1. LikelyTogether AI spends the round mainly on data center and chip capacity, not new products.
  2. Possiblemore sovereign wealth funds, like UAE or Qatar, make similar AI infrastructure bets within a year.
  3. Possiblea closed-model lab cuts inference prices to compete with open-weight economics.
  4. Wild CardTogether AI itself goes public within 18 months, riding the neocloud wave.

🥄 The Spoon Take

Saudi oil money just placed a bet on open-weight AI over the closed labs. That's a signal, not just a check. The smartest capital in the world increasingly treats 'run any open model cheaply' as the safer bet than one lab's frontier model.

🤔 Pushback

Aramco leading may just mean sovereign funds got priced out of OpenAI and Anthropic rounds, not that open models are winning.

·TechCrunch - $800M Series C raise·BusinessWire - Official funding announcement·TechTimes - Open models break $1B
Thursday Jul 2
FUNDINGOther

Chamath Raises $135M, Becomes CEO

2JUL
8090 LABSCHAMATH$135M

An All-In podcast host just bet big on himself again. Chamath Palihapitiya raised $135M for his AI coding startup and became CEO. Salesforce Ventures led the round.

8090 Labs builds 'Software Factory,' an AI coding agent for corporate programming teams. It targets healthcare, insurance, aerospace, and government work Cursor and Claude Code rarely touch.

Backers include Jeffrey Katzenberg, David Sacks, and fellow All-In host David Friedberg. Palo Alto Networks CEO Nikesh Arora chipped in too. Palihapitiya steps back into full-time operating work for the first time since Facebook.

He compares this AI moment to social media's early rise at Facebook. Audit trails and enterprise controls are 8090's pitch to the industries everyone else avoids.

full brief & sources

⚡ Why this matters

  • A marquee VC name jumping back into full-time operating work is rare.
  • 8090 targets the regulated industries most AI coding tools ignore.
  • Salesforce Ventures backing signals where enterprise AI coding money is headed next.

🔍 What happened

  • 8090 Labs closed a $135 million Series A on June 29, led by Salesforce Ventures.
  • Palihapitiya founded 8090 Labs in January 2024; he now takes the CEO seat.
  • Backers include Jeffrey Katzenberg's WndrCo, David Sacks' Craft Ventures, and All-In co-hosts David Friedberg and Jason Calacanis.
  • Angel investors include Palo Alto Networks CEO Nikesh Arora and Quora CEO Adam D'Angelo.
  • The product, Software Factory, adds audit trails and enterprise controls to AI-built code.
  • Target customers: healthcare, insurance, aerospace, energy, financial services, and government.

💬 Smart takes

  • Chamath Palihapitiya: 'I was waiting for a moment like this to return to a full-time operating role.'
  • Skeptic: a founder-turned-investor-turned-founder-again story is also a great way to re-inflate a personal brand.

🧭 Where this goes

  1. Likely8090 leans hard on 'enterprise-grade' and 'audit trail' messaging against Cursor and Claude Code.
  2. LikelyPalihapitiya uses the All-In podcast as a free, weekly marketing channel for 8090.
  3. Possiblea regulated-industry customer win becomes 8090's proof point within two quarters.
  4. Wild Card8090 gets acquired by a larger enterprise vendor once it proves the regulated-industry niche.

🥄 The Spoon Take

Every VC calls AI coding the biggest market in a decade. Chamath put his own job on that bet, not just his firm's money. The wager isn't the model, it's the compliance paperwork around it. Regulated industries are the boring part of software that actually pays.

🤔 Pushback

Founders who return as 'visionary CEO' after years away from operating often struggle with the daily grind that made the first company work.

·TechCrunch - $135M raise, Chamath CEO·Startup Fortune - 8090 Labs details·MLQ News - Palihapitiya CEO move
Saturday Jun 27
FUNDINGOpenAI

OpenAI May Push Its IPO To 2027

27JUN
RACE OFFFILED FIRSTWAITS 2027

The sprint to list first just reversed. OpenAI is weighing a wait until 2027, per June reports. Its rival already filed. Altman is holding out for a trillion-dollar price.

For months the narrative was a dash to go public. SpaceX debuted in June. Anthropic submitted draft paperwork on the first. Now the ChatGPT maker may push its market debut to 2027.

The holdup is the number. Altman has leaned on bankers for a headline valuation near a trillion. The last private deal valued the company around $730 to $852 billion. Public buyers may balk at that leap.

The pressure order changes. Anthropic now meets the market first. Every margin and customer figure it reveals becomes the yardstick for OpenAI. Loudest in the hype cycle is not first to the bell.

full brief & sources

⚡ Why this matters

  • Reverses the 2026 narrative that OpenAI would lead the AI IPO wave.
  • Puts Anthropic, which already filed, first to face public scrutiny.
  • Signals public markets may balk at trillion-dollar AI valuations.

🔍 What happened

  • June 25 reports say OpenAI is considering delaying its IPO to 2027.
  • It had eyed a Q3 or Q4 2026 debut earlier.
  • Altman is pushing for a near $1 trillion valuation.
  • OpenAI's last private round valued it between $730 and $852 billion.
  • Anthropic filed its draft S-1 on June 1 after a $965 billion round.

💬 Smart takes

  • Reports (June 25): Altman has fiercely pushed advisers to reach a $1 trillion valuation.
  • Skeptic: 'considering a delay' is not a decision, and this could be negotiating leverage, not a real retreat.

🧭 Where this goes

  1. LikelyAnthropic's filed numbers become the benchmark every AI lab is judged against.
  2. PossibleOpenAI raises another huge private round to avoid listing early.
  3. Possibleappetite for AI IPOs cools if early listings trade flat.
  4. Wild CardOpenAI never IPOs at $1T and restructures as something other than a normal public company.

🥄 The Spoon Take

Going public second is not losing. It is letting someone else absorb the first hit. Anthropic now sets the comp for AI lab economics with real disclosed numbers. If those numbers look shaky, every valuation in the field resets. OpenAI gets to watch before it jumps.

🤔 Pushback

A reported 'considering' is cheap, and Altman has reversed course before, so this may be leverage with bankers rather than a real delay.

·Yahoo Finance - IPO delay report·Built In - 2026 IPO watchlist
Monday Jun 15
FUNDINGOther

Wall Street Now Finances AI Compute

15JUN
$35B 20 GIGAWATTSAI COMPUTECAPITAL

Compute just became something Wall Street can finance. Apollo and Blackstone are leading a $35 billion deal to build 20 gigawatts of Broadcom AI chips for Anthropic and OpenAI.

AI infrastructure used to be a hyperscaler capex line. Now it's an asset class.

Apollo, Blackstone, and global banks bundled chips, networking, power, and customer commitments into one financed platform. It targets over 20 gigawatts through 2028. Anthropic's 1-gigawatt expansion starts mid-2026.

Apollo's Jamshid Ehsani called it the largest private financing ever. Compute now gets funded like toll roads.

full brief & sources

⚡ Why this matters

  • Compute is the binding limit on every AI roadmap. This is how it gets funded at scale.
  • Private credit, not just hyperscaler balance sheets, now underwrites frontier compute.
  • An asset-backed structure lets investors finance AI buildout like infrastructure.

🔍 What happened

  • Jun 9: Apollo leads a $35B capital solution for Broadcom's AI XPV Platform.
  • Partners: Blackstone's credit arm plus a syndicate of global banks.
  • Goal: over 20 gigawatts of Broadcom XPU compute for frontier labs through 2028.
  • Named customers: Anthropic and OpenAI; Anthropic's 1GW expansion starts mid-2026.
  • Apollo's Jamshid Ehsani: 'the largest private financing ever executed.'

💬 Smart takes

  • Apollo (Jamshid Ehsani): 'the largest private financing ever executed.'
  • Blackstone: framed it as a landmark platform to accelerate 20+ gigawatts of AI deployment.
  • Skeptic: asset-backed compute assumes demand holds; a model-demand air pocket leaves lenders with depreciating chips.

🧭 Where this goes

  1. Likelymore chip-plus-power-plus-offtake bundles get financed by private credit in 2026.
  2. LikelyBroadcom's custom XPU path gains share against merchant GPUs for labs.
  3. Possible'gigawatts financed' becomes a standard AI-infrastructure metric.
  4. Wild Carda compute-backed security gets securitized and traded like mortgage bonds.

🥄 The Spoon Take

This is the moment AI compute stops being a tech cost and becomes an asset class. Bundling chips, power, and customer contracts into financeable platforms is how railroads and pipelines got built. The constraint on AI was never just chips. It was who pays for the buildout. Now Wall Street does.

🤔 Pushback

Asset-backed compute only works if demand holds; if model revenue stalls, lenders are left holding 20 gigawatts of depreciating silicon.

·Apollo - $35B compute platform·Blackstone - 20GW AI platform·GlobeNewswire - $35B capital solution
Sunday Jun 14
FUNDINGOther

Bezos Raises $12B For Physical-World AI

14JUN
BUILDS STUFF$12B

Jeff Bezos just put real money behind AI for the physical world. His startup Prometheus raised $12 billion at a $41 billion valuation. The goal: an 'artificial general engineer' for cars, chips, and drugs.

Bezos co-leads Prometheus with Stanford scientist Vik Bajaj. The two showed up together on CNBC for the first time. Total funding now tops $18 billion.

Most AI money chases chatbots and code. Prometheus wants AI that designs and builds real things. Targets: aerospace, automotive, manufacturing, drug discovery.

It has about 150 people and no public product yet. Backers include JPMorgan, BlackRock, and Goldman Sachs. They are betting on the founder, not a demo.

full brief & sources

⚡ Why this matters

  • It's one of the largest early-stage rounds ever for a company with no shipping product.
  • Bezos is steering AI away from chatbots toward physical engineering and manufacturing.
  • A $41B valuation on 150 people sets a new bar for founder-led AI bets.

🔍 What happened

  • Prometheus raised $12 billion in a Series B at a roughly $41 billion valuation, announced June 11.
  • Co-CEOs are Jeff Bezos and Stanford scientist Vik Bajaj.
  • With its $6.2 billion launch round, total funding tops $18 billion.
  • Backers include JPMorgan, BlackRock, Goldman Sachs, DST Global, and Arch Venture.
  • The aim is an 'artificial general engineer' for computing, aerospace, automotive, and drug discovery.

💬 Smart takes

  • Bezos: 'We're not being secretive,' he told CNBC about the long-quiet startup.
  • Axios: Prometheus is now worth $41 billion building 'industrial AI.'
  • Skeptic: $41B for 150 people and no product is a bet on Bezos, not results.

🧭 Where this goes

  1. LikelyPrometheus stays quiet on product specifics through 2026.
  2. Possibleit lands a marquee manufacturing or aerospace pilot within 12 months.
  3. Possiblethe 'physical AI' label pulls more capital into robotics and materials startups.
  4. Wild Cardan 'artificial general engineer' demo reshapes how hardware gets designed by 2028.

🥄 The Spoon Take

The smart-money bet is shifting from screens to atoms. Chatbots and coding agents are crowded. Bezos is buying into physical-world AI, design, materials, manufacturing, where moats run deeper and rivals are fewer. $12 billion with no product says investors are pricing the thesis, not the demo.

🤔 Pushback

Physical-world AI is far harder than software, and 'artificial general engineer' could stay a pitch deck for years.

·TechCrunch - Prometheus $12B raise·GeekWire - $41B valuation·CNBC - Bezos interview
FUNDINGOther

China's EngineAI Files To Go Public

14JUN
PUBLIC10,000 BOTSHONG KONG IPO

China's humanoid robot race hit the public markets. EngineAI, a Shenzhen maker shipping 10,000 robots a year, filed for a Hong Kong IPO. It joins a wave of Chinese robot makers chasing billions.

EngineAI was founded in 2023. In June it opened a 12,000 square meter Shenzhen factory and started shipping its T800 robots.

The line can build a humanoid every 15 minutes, geared for 10,000 units. The pitch to investors is real production, not viral demo clips.

It is not alone. Unitree filed for a $7 billion IPO. PaXini, Dreame, and Linkerbot are all circling Hong Kong listings.

full brief & sources

⚡ Why this matters

  • First wave of Chinese humanoid makers reaching public markets.
  • Production volume, not demos, is now the pitch to investors.
  • China is building a domestic robotics IPO pipeline to rival US AI listings.

🔍 What happened

  • EngineAI filed confidentially for a Hong Kong IPO, reported June 12, 2026.
  • Working with China International Capital Corp and Citic Securities.
  • Raised about $27.9M Series B in April 2026 at a ~$1.5B valuation.
  • Opened a 12,000 sqm Shenzhen factory on June 1, shipping T800 robots.
  • Line can produce one humanoid every 15 minutes, geared for 10,000 units.
  • Founded in 2023; went viral with a robot front-flip clip.

💬 Smart takes

  • The Next Web: the pitch is a real manufacturing base, not viral demo clips.
  • Bloomberg: EngineAI is working with CICC and Citic Securities on the listing.
  • Skeptic: billion-dollar valuations on under two years of operating history echo classic bubble setups.

🧭 Where this goes

  1. Likelymultiple Chinese humanoid makers complete Hong Kong listings in 2026.
  2. LikelyWestern robotics startups cite the China pipeline to raise their own rounds.
  3. Possiblea Chinese humanoid maker posts real industrial-deployment revenue this year.
  4. Wild Carda high-profile humanoid IPO breaks badly and chills the whole sector.

🥄 The Spoon Take

The humanoid story is moving from demo videos to factory floors and order books. China is industrializing it fast and taking it public faster. The question for everyone else is whether real production volume shows up before the valuations need defending.

🤔 Pushback

A factory that can build 10,000 robots is not the same as 10,000 robots sold. Demand, not capacity, decides if these valuations hold.

·The Next Web - EngineAI HK IPO·Bloomberg via Investing - EngineAI filing·CryptoBriefing - EngineAI IPO
Wednesday Jun 10
FUNDINGOpenAI

OpenAI Files for IPO After Anthropic

10JUN
$1T DEBUT?OPENAI$1T?

The entire frontier AI industry is going public at once. OpenAI confidentially filed for IPO on June 8, targeting a $1T+ listing this fall. SpaceX at $1.8T, Anthropic at $965B, OpenAI at $852B - all listing in 2026.

Sam Altman spent two years saying OpenAI would stay private. The Microsoft revenue cut and the compute cost of o-series models changed the timeline.

What this means operationally: quarterly earnings creates pressure to ship faster and treat safety investments as cost centers with measurable ROI.

The tell - Dario publicly called for an AI pause the same week Anthropic filed. Every board meeting now has a fiduciary dimension attached to existential-risk decisions.

full brief & sources

⚡ Why this matters

  • IPO changes AI lab decision-making permanently - every safety call, hiring freeze, and pricing move now has a quarterly earnings context attached
  • Three frontier labs going public simultaneously is an unprecedented concentration of AI valuation entering public markets
  • Sam Altman spent two years saying OpenAI would stay private - the timeline shift signals something changed: the Microsoft restructuring, compute costs, or investor pressure

🔍 What happened

  • OpenAI filed confidentially with the SEC on June 8 - one week after Anthropic filed at $965B implied valuation
  • Target window: fall 2026 listing at $852B implied valuation based on secondary market pricing
  • SpaceX ($1.8T), Anthropic ($965B), and OpenAI ($852B) are all in the 2026 pipeline - combined implied market cap exceeds $3.6T

💬 Smart takes

  • Dario Amodei called for an AI pause the same week Anthropic filed - a preview of every lab CEO's new reality: existential safety arguments in the same breath as investor roadshows
  • The Microsoft restructuring (reduced OpenAI revenue share) is the rumored catalyst that accelerated the filing
  • Confidential filings allow a company to test investor appetite before committing to a public prospectus

🧭 Where this goes

  1. Bloomberg, The Verge, Reuters - June 8 2026

🥄 The Spoon Take

The IPO gate changes AI lab culture permanently. Quarterly earnings mean quarterly pressure on margins, releases, and headcount. Dario calling for an AI pause the week Anthropic filed is the new norm. Every frontier lab decision now has a Wall Street read attached.

🤔 Pushback

Confidential filings aren't IPOs - all three could pull back if markets turn or regulatory hurdles emerge before the listing window.

·Bloomberg - OpenAI confidential IPO filing·The Verge - AI IPO wave 2026·Reuters - OpenAI IPO details
Tuesday Jun 9
FUNDINGOther

SpaceX Prices Thursday. Anthropic Watches.

9JUN
$1.75TANTHROPICOPENAI

The world's largest IPO prices Thursday, and Anthropic has a financial stake. SpaceX targets $75B at a $1.75T valuation, listing as SPCX on Nasdaq. The result will set the ceiling for Anthropic and OpenAI's own debuts.

Goldman Sachs leads a 21-bank underwriting syndicate. Thirty percent of the float is reserved for retail investors. The raise would be the largest in recorded IPO history.

The compute connection: the company pays $1.25B/month for Colossus, the GPU cluster originally built for xAI. If pricing comes in weak, Colossus expansion slows and Claude's capacity plans get pinched.

Watching this one matters because it's the first trillion-dollar AI-adjacent listing of the public-market era. Claude's maker filed its S-1 on June 1, OpenAI signals a 2026 path. Thursday's reception influences both.

full brief & sources

⚡ Why this matters

  • SPCX is the first public market test of a trillion-dollar AI-era valuation - its reception tells Anthropic and OpenAI whether institutional investors will absorb two more massive AI equity offerings this year
  • Anthropic's compute dependency on SpaceX Colossus ($1.25B/month) means SpaceX's IPO-funded expansion plans directly affect Anthropic's capacity roadmap
  • For any PM building on Claude, the financial health of the AI infrastructure layer (SpaceX Colossus, CoreWeave, etc.) determines API availability and pricing stability

🔍 What happened

  • SpaceX IPO pricing: June 11, Nasdaq, ticker SPCX, targeting $75B raise at $1.75T valuation
  • Goldman Sachs leads a 21-bank underwriting syndicate; roadshow completed over the past two weeks
  • 30% of the float allocated to retail investors via Robinhood, Fidelity, and Schwab - unusually high retail allocation for a mega-IPO
  • SpaceX's AI division (xAI, Grok) generated $14B in cash consumption in 2025 against $3.2B in revenue - Starlink's operating profit subsidizes the model business
  • Anthropic filed its S-1 on June 1, 2026. OpenAI has signaled public market readiness by end of 2026
  • The SPCX IPO result sets the public market comp for both - if SPCX prices 20%+ above range, it validates trillion-dollar AI valuations; if it prices at the bottom, it introduces doubt

💬 Smart takes

  • Goldman Sachs (lead underwriter): Institutional demand described as 'significantly oversubscribed' across the US, Europe, and Asia Pacific tranches as of close of roadshow June 8.
  • Skeptic read: xAI is losing $14B/year on $3.2B in revenue. If institutional investors price that segment skeptically (as they did WeWork's ancillary businesses), the blended valuation could land well below $1.75T range, creating a read-through problem for Anthropic's S-1 multiples.

🧭 Where this goes

  1. LikelySPCX prices within range on Thursday and opens up 10-15% Friday, validating mega-cap AI-adjacent IPO appetite and easing Anthropic's path
  2. LikelyAnthropic sets its IPO roadshow for Q3 2026 using SPCX's price-to-ARR multiple as the floor comp
  3. PossibleRetail demand (30% of float through Robinhood/Fidelity) outpaces institutional, causing a first-day pop that then fades - creating short-term noise but long-term validation
  4. Wild CardxAI's losses trigger institutional haircut requests that push SPCX pricing below the $1.65T floor, forcing Goldman to activate stabilization mechanisms and delaying Anthropic's S-1 process by 60-90 days

🥄 The Spoon Take

This is the IPO that answers the question everyone in AI is asking: will public markets absorb trillion-dollar AI valuations? Anthropic and OpenAI are watching SPCX pricing the way a startup watches a category leader's Series B. If SPCX flies, the AI IPO era begins. If it stumbles, the queue backs up.

🤔 Pushback

SpaceX's Starlink unit is a cash-generating infrastructure business unlike Anthropic's pure-software model - its IPO success doesn't directly de-risk Anthropic's valuation thesis.

·BFWAI - AI News Today June 8 2026·Stratechery - The iPhone's Last Stand (SpaceX context)
Monday Jun 1
FUNDINGAnthropic

Anthropic Files Before OpenAI

1JUN
CLAUDES-1 FILED

Anthropic confidentially filed its draft S-1 with the SEC on June 1, four days after closing a $65B Series H at a $965B post-money valuation. The filing lands ahead of OpenAI's expected fall listing, putting Anthropic on track to be the first frontier AI lab to trade publicly.

First frontier AI lab to formally enter the IPO path. $965B private valuation, $47B revenue run-rate, up from $9B at end of 2025 — a 5x jump in six months.

Anthropic filed BEFORE OpenAI. Both labs had been telegraphing fall 2026 timing; Dario moved the timeline up after the Series H closed Wednesday. If Anthropic debuts at $1T it would rank as the 2nd or 3rd largest IPO ever (behind SpaceX, Saudi Aramco). And yes — Claude almost certainly wrote the S-1.

Frontier-lab valuations become market-priced, not VC-priced. Compute partners, customers, and talent decisions will shift as Anthropic's stock becomes the daily public read on what AI is worth.

full brief & sources

⚡ Why this matters

  • First frontier AI lab to formally file for public market listing
  • Beats OpenAI to the SEC by ~4 months (OpenAI was expected to file in fall 2026)
  • A trillion-dollar AI IPO would rank as the 2nd or 3rd largest in history (behind SpaceX and Saudi Aramco)

🔍 What happened

  • June 1, 2026: Anthropic confidentially filed draft S-1 with SEC
  • Comes 4 days after Series H closed at $965B post-money valuation
  • Revenue run-rate $47B, up from $9B at end of 2025 (5x growth in 6 months)
  • Filing remains confidential while SEC reviews; Anthropic chooses public timing after
  • Anthropic statement: "Gives us the option to go public after the SEC completes its review"
  • OpenAI reportedly preparing its own filing for fall 2026 — Anthropic just pulled the trigger first

💬 Smart takes

  • Anthropic (official statement): "The proposed IPO will depend on market conditions and other factors"
  • CNBC: Frames the filing as Anthropic "prepping Wall Street for landmark AI deal"
  • The Register: Headlines it as Anthropic "now atop the AI bubble"
  • Skeptic — Ed Zitron (ongoing): Per his "Wheresyourdata" thesis, the $47B run-rate depends heavily on prepaid compute deals and stock-based compensation. Public-market scrutiny will pressure-test the revenue quality in a way private rounds never did

🧭 Where this goes

  1. If SEC review is fast: Anthropic could trade as early as Q3 2026; Q4 more likely
  2. OpenAI accelerates its own filing to avoid being seen as the slower mover
  3. Frontier-lab valuations get re-rated every market open, not every VC round — daily public price discovery for AI
  4. Compute partners (Google, AWS, Broadcom) now hold shares of a publicly traded customer they're contractually entangled with
  5. AI labor market shifts as Anthropic equity becomes liquid — Anthropic poaching gets dramatically easier

🎯 Implication

  • For PMs: the AI vendor you're building on now has public-market accountability. Expect more transparency around revenue mix, customer concentration, churn — plus more PR optics on every enterprise win
  • For execs: re-evaluate dependency risk on Anthropic now that it's subject to short-seller pressure, activist investors, and quarterly earnings cycles. The CRM-headless thesis applies to Anthropic itself now
  • For founders: AI-native startup valuations get a public-market multiple to benchmark against — could re-rate up (Anthropic trades at premium) or down (S-1 reveals soft enterprise pull-through). Either way, the era of "AI valuations only make sense to VCs" is ending
·Anthropic - Confidential S-1 filing·CNBC - IPO prospectus filed·TechCrunch - Files to go public·The Register - Atop the AI bubble
Saturday May 30
FUNDINGOther

Groq's $650M Second Act

29MAY
GROQ$650M

Groq is raising $650M from existing backers to relaunch as an inference neocloud. In December, Nvidia paid $20B in a "not-acqui-hire" that took Groq's top engineers and licensed its chip tech. Interim CEO Adam Winter and CFO Matt Eng now lead the rebuild. Existing investors Disruptive and Infinitium agreed to backstop the round.

Nvidia bought the team and the IP for $20B. The shell raises $650M to keep going. The market structure here is new.

The $650M is effectively guaranteed: Disruptive and Infinitium committed to fill any pro-rata shortfall. Groq is pivoting from chip-vendor to inference-as-a-service, going head-to-head with CoreWeave, Together, and Lambda. The new Groq sells inference - on the chips it licensed away to Nvidia. Strange new shape: a competitor running on a competitor's licensed tech.

For PMs evaluating inference vendors: Groq's customer continuity is a real risk. For execs: this is what "Nvidia consolidates" looks like when antitrust forbids the full acquisition. For investors: every other specialty silicon startup gets repriced against the Groq-Nvidia structure now.

full brief & sources

⚡ Why this matters

  • Nvidia paid $20B in December 2025 to take Groq's engineers and license the tech without triggering antitrust. That structure is now the template.
  • Groq is pivoting to inference neocloud - the second act for an Nvidia challenger that lost its talent.
  • Inference compute is now bigger than training compute. The neocloud category (CoreWeave, Together, Lambda) is where the dollars are flowing.

🔍 What happened

  • May 28, 2026: Axios scoops Groq raising $650M from existing investors.
  • May 29: TechCrunch confirms; Yahoo Finance, Seeking Alpha, The Next Web pick it up.
  • Existing backers Disruptive Ventures and Infinitium have agreed to backstop the round if other investors decline pro-rata.
  • Leadership: Adam Winter (interim CEO), Matt Eng (CFO). The founding team left for Nvidia in the December $20B deal.
  • Strategy: pivot from selling LPU chips to running an inference cloud service powered by them.
  • Inference compute now larger than training compute, per industry sources.

💬 Smart takes

  • The Next Web framing: "Nvidia paid Groq $20 billion and took its top engineers. Now Groq is raising $650 million for what's left."
  • Axios scoop framing: "Groq's second act" - the existing backers are deciding it's worth funding the remnant.
  • Skeptic: A neocloud is a low-margin commodity business. Without the founding chip team, what's the moat against Together AI or Lambda?

🧭 Where this goes

  1. Round closes at $650M before end of Q2 2026, with a strategic Nvidia or CoreWeave reseller deal layered in.
  2. Other specialty silicon startups (Cerebras, Etched, MatX) get the Nvidia "licensing buyout" pitch within 12 months.
  3. Antitrust regulators (FTC, DOJ, EU) launch reviews on the not-acqui-hire structure by Q3.
  4. Inference neocloud category sees $5B+ in fresh funding across Together, CoreWeave, Lambda, Groq by end of 2026.

🎯 Implication

  • For PMs evaluating inference vendors: price Groq into your stack only if you can switch within 30 days. Customer continuity risk is high.
  • For execs: watch the not-acqui-hire pattern - Nvidia's $20B move is the new playbook for sidestepping antitrust on AI hardware.
  • For investors: the floor on chip startups is now "what would Nvidia pay to license you out of the market." That's the new comp.
·TechCrunch - Groq $650M raise·Axios - Second act scoop·The Next Web - $20B for what's left·Seeking Alpha - $650M fundraise
Friday May 29
FUNDINGAnthropic

Anthropic Tops OpenAI At $965B

28MAY
$965 BILLION!12$$ANTHROPICOPENAI

Anthropic raised $65 billion at a $965 billion post-money valuation, eclipsing OpenAI ($852B) for the first time. Memory chip giants Samsung, SK Hynix, and Micron joined as strategic infrastructure partners. Run-rate revenue crossed $47B this month. IPO expected this autumn.

2.5x valuation jump in 90 days. $380B in February. $965B now. Three months. The number that matters most: Anthropic just passed OpenAI by $113B.

Altimeter, Dragoneer, Greenoaks, and Sequoia each wrote checks over $2B. Google pledged up to $40B over time. Amazon added $5B. The new twist is the memory tier joining the cap table. Samsung, SK Hynix, and Micron together represent the world's HBM supply. They are now financially aligned with Anthropic's compute roadmap, not just selling into it.

For PMs: the lab-stability risk on Claude bets just dropped again. For execs: assume Anthropic and OpenAI both file public S-1s before Q4. For procurement: enterprise pricing power now lives with the lab that's profitable, not the one that's biggest.

full brief & sources

⚡ Why this matters

  • First time Anthropic's valuation has passed OpenAI's. The market called the lead change.
  • Memory chip vendors joining the cap table is a new shape. HBM supply is now strategically aligned to Claude.
  • Last private round before IPO. The public-market comp for AI labs gets set in the next 6 months.

🔍 What happened

  • May 28, 2026. Anthropic closes $65B Series H at $965B post-money.
  • Co-leads: Altimeter Capital, Dragoneer, Greenoaks, Sequoia Capital. Each over $2B.
  • Strategic infrastructure partners: Samsung Electronics, SK Hynix, Micron Technology. First time all three memory giants joined a single AI round.
  • Institutional: Baillie Gifford, Blackstone, Brookfield, Coatue, D1 Capital, D.E. Shaw Ventures, DST Global, Fidelity, Capital Group.
  • Hyperscalers: $5B from Amazon (April commitment), Google pledged up to $40B over time.
  • Run-rate revenue: crossed $47B earlier in May.
  • Prior round: $30B at $380B post-money in February 2026. 2.5x jump in 90 days.
  • OpenAI last priced at $852B in February. Anthropic now ahead by $113B.
  • Expected IPO timing: autumn 2026.

💬 Smart takes

  • Dario Amodei (Anthropic CEO, statement): "advance our safety and interpretability research, expand compute to meet growing demand for Claude, and scale the products and partnerships our customers rely on."
  • Korea Herald on the memory deal: SK Hynix and Micron's participation is a defensive move. They get early sight into next-gen HBM bandwidth requirements before competitors.
  • SamMobile: Samsung is the only one of the three with foundry capacity. Anthropic-Samsung scope may extend beyond memory into chip manufacturing.
  • Skeptic: $965B at $47B ARR is 20x revenue. OpenAI sits at $852B with stated 2026 revenue near $20B (≈43x). Anthropic is cheaper on revenue but only because OpenAI got priced last. The IPO will reset both.

🧭 Where this goes

  1. Both labs file public S-1s before Q4 2026.
  2. Memory pricing for HBM4E and beyond gets co-designed with Anthropic's roadmap before Q2 2027.
  3. OpenAI counters with a similarly-sized round at $1T+ valuation by July.
  4. By 2027, the public-market comp on AI labs sits between 15x and 25x ARR. Below that and one lab gets bought.

🎯 Implication

  • For PMs building on Claude: vendor-stability risk just dropped. Lock multi-year API commits at current pricing.
  • For execs choosing AI vendors: the dual-IPO comp means both labs need to keep revenue compounding. Negotiate from a buyer's market through Q3.
  • For investors tracking AI: the memory-chip strategic investor pattern repeats with Samsung Foundry on a major lab deal within 12 months.
·Anthropic - Series H announcement·Bloomberg - $965B eclipses OpenAI·TechCrunch - Nears $1T pre-IPO·Korea Herald - Memory giants invest
Thursday May 28
FUNDINGOther

Cognition Buys Its Independence

27MAY
C$1B RAISED!NO BIG TECH OVERLORDCOGNITION / DEVIN

Cognition, maker of autonomous coding agent Devin and acquirer of Windsurf, raises $1 billion at a $26 billion post-money valuation. Lux Capital, General Catalyst, and 8VC co-lead. CEO Scott Wu uses Bloomberg TV to flag the SpaceX-Cursor deal and pick independence.

$26B post-money. 2.5x jump from $10.2B last September. Run-rate revenue at $492M. Up 13x in 12 months.

Devin now writes more than 90% of Cognition's own code. Enterprise customers: Goldman Sachs, Citi, Mercedes-Benz, US Army, US Navy. Wu went on Bloomberg TV to name the SpaceX-Cursor $60B option and frame this round as the independent path. The coding-agent consolidation is on. Cursor (Anysphere) has the SpaceX option. Lovable cleared App Store. Cognition just funded survival.

For PMs picking coding tools: bake-off Devin against Claude Code and Codex this quarter. For execs: assume AI-coding M&A premiums get richer through 2026. For founders in agent-adjacent spaces: pick a side - absorbed or independent - before the round closes around you.

full brief & sources

⚡ Why this matters

  • Largest coding-agent funding round of 2026 by valuation jump.
  • Wu explicitly framed independence vs the SpaceX-Cursor template. First named refusal.
  • 13x revenue growth in 12 months is the kind of curve that gets enterprises to ditch incumbents.

🔍 What happened

  • May 27, 2026. Cognition closes $1B at $26B post-money.
  • Co-leads: Lux Capital, General Catalyst, 8VC.
  • Participation: Founders Fund, Ribbit Capital, Atreides Management.
  • Previous round: $10.2B in September 2025. 2.5x jump in 8 months.
  • Run-rate revenue: $492M (up from $37M in May 2025).
  • Enterprise customer names: Goldman Sachs, Citi, Mercedes-Benz, US Army, US Navy.
  • More than 90% of Cognition's own internal code is now written by Devin.
  • Acquired Windsurf earlier in 2026 to bundle agent + IDE.

💬 Smart takes

  • Scott Wu (Cognition CEO, Bloomberg TV): the raise keeps Cognition independent, a pointed comment given the SpaceX-Cursor deal.
  • Lux Capital (lead): revenue compounding 13x in 12 months is the trigger to back independence over consolidation.
  • Skeptic: $492M ARR at $26B valuation is 53x revenue. Cursor at $2B ARR went to $60B option (30x). Either Cognition is the next acquisition target at a higher multiple, or it has to grow into the price.

🧭 Where this goes

  1. Cursor and Cognition compete head-to-head for enterprise dev budgets through Q4.
  2. Anthropic answers with Claude Code Enterprise tier and a marketplace SDK by Q3.
  3. Two more coding-agent fundings or M&A events by end of summer.
  4. By 2027, the coding-agent layer consolidates to 3-4 winners. Cognition, Cursor, Claude Code, Codex are the candidates.

🎯 Implication

  • For PMs picking coding tools: bake-off Devin against Claude Code and Codex this quarter. The capability and price gap is closing.
  • For execs: assume AI-coding tools become a four-vendor RFP by end of year.
  • For founders in agent-adjacent categories: get clarity on absorbed vs independent before your next round prices.
·Bloomberg - Cognition $1B raise·TechCrunch - $25B pre-money round·TNW - 90% AI-written code·The Decoder - Devin valuation doubled
Wednesday May 27
FUNDINGMistral

Mistral Hits $50B Valuation

27MAY
M$50 BILLION!!!$$$$MISTRAL

Mistral (the Paris-based open-weights AI lab founded by ex-Meta and ex-DeepMind researchers) raises a fresh round at $50B. First European AI lab past the $50B mark. Sequoia leads, with Saudi PIF and Lightspeed joining.

Mistral closed a fresh round at a $50B valuation. 5x its last mark in mid-2025. Largest European AI lab funding round ever.

The cap table now shows Sequoia (lead), Saudi PIF, Lightspeed, Andreessen Horowitz, and General Catalyst. PIF and Sequoia each wrote checks above $2B. Existing investors maintained pro-rata.

Mistral is doubling down on open-weights as the wedge against OpenAI/Anthropic. Plans to release Mistral Large 3 within 90 days as open-weights under Apache 2.0.

full brief & sources

⚡ Why this matters

  • First European AI lab past $50B - changes the geography of AI capital.
  • Validates open-weights as a viable commercial bet vs closed labs.
  • Sovereign capital (PIF) entering at scale signals nation-state AI competition heating up.
  • Sequoia returning at this size after smaller earlier rounds means they think Mistral wins or comes close.

🔍 What happened

  • Round size: ~$2.5B at $50B post-money.
  • Lead: Sequoia. Co-investors: Saudi PIF, Lightspeed, a16z, General Catalyst.
  • Existing French strategic LPs (Bpifrance, CMA CGM) maintained pro-rata.
  • Use of funds: GPU buildout, Mistral Large 3 (Apache 2.0 weights), enterprise GTM in EMEA.
  • Headcount target: 600 by end of 2026 (currently ~400).

💬 Smart takes

  • Sequoia (Roelof Botha): Mistral is the open-weights bet that scales globally. Best technical team in Europe.
  • Skeptic: Open-weights commercial economics still unproven vs OpenAI's $5B+ ARR. $50B is a big price for a #4 player.
  • French government: Mistral is now formally a national AI champion. Expect EU AI Act influence to follow.
  • OpenAI execs (off the record): Mistral has 18 months to prove enterprise ARR or this round looks rich.

🧭 Where this goes

  1. Mistral Large 3 launches Q3 2026 (open-weights Apache 2.0).
  2. European sovereign cloud partnerships expand (OVH, Scaleway).
  3. Saudi PIF pushes Mistral into Middle East government / energy verticals.
  4. Closed labs (OpenAI/Anthropic) double down on enterprise lock-in via private models + agents to widen the gap.

🎯 Implication

  • For PMs: Mistral's open-weights are now a credible enterprise build option for cost-sensitive teams. Add to vendor matrices.
  • For execs: European AI procurement shifts. Data sovereignty pitch (Mistral runs on EU soil) becomes harder for closed labs to counter.
  • For builders: $50B sets a new ceiling for open-weights lab valuations. Expect more open-weights labs (Together, Reka, etc.) to raise at higher prices.
  • For investors: Sovereign capital + commercial VC mixing at $50B+ rounds is the new normal. Solo-VC rounds at this size are over.
·FT - Mistral closes $2.5B at $50B valuation·Bloomberg - Saudi PIF joins Mistral cap table·TechCrunch - Sequoia leads Mistral mega-round·Les Echos - La French Tech a son hectacorne IA
FUNDINGOther

Inference Routing Hits $113 Million

26MAY

AI inference routing company OpenRouter raises $113 million Series B led by CapitalG (Google's growth fund). The model-routing layer is becoming an investable category.

OpenRouter is a multi-LLM router. You send a request, it picks which provider answers based on price, latency, or quality.

CapitalG leading is significant. That's Google's growth fund backing a company that explicitly arbitrages Google's own Gemini against Anthropic, OpenAI, Mistral, and others. Either Google is hedging or it doesn't see OpenRouter as a competitor.

If LLM-recall and price-per-token become the dominant procurement axes, routers become essential. Watch Portkey and LiteLLM follow.

full brief & sources

⚡ Why this matters

  • The model-routing layer just got institutional validation.
  • CapitalG (Google's growth fund) backing a multi-provider router signals the routing category is durable.
  • Cross-vendor abstraction is becoming infrastructure, not a workaround.

🔍 What happened

  • May 26, 2026. OpenRouter raises $113M Series B.
  • Round led by CapitalG (Google's growth-stage fund).
  • OpenRouter routes API requests across Anthropic, OpenAI, Google, Mistral, DeepSeek, and dozens of open-weight providers.
  • Pricing model: pass-through token costs plus a thin routing margin.
  • Founded 2023. Previously raised seed and Series A from a16z and others.

💬 Smart takes

  • OpenRouter: the router is the abstraction layer enterprises will need as model diversity grows.
  • CapitalG (Google's growth fund) leading: notable that Google's own growth arm backs a router that arbitrages Gemini against competitors.
  • Skeptic: if model APIs converge on a standard (OpenAI's API shape is already de facto), routing becomes a commodity. The moat is reliability and observability, not capability.

🧭 Where this goes

  1. Portkey, LiteLLM, and other routing players raise comparable rounds within 6 months.
  2. Enterprise procurement starts asking 'do you use a router or call APIs directly?' as a vendor diligence question by Q4.
  3. Hyperscalers respond with native routing in Bedrock, AI Foundry, Vertex (Azure already partial).
  4. OpenRouter starts pushing into governance and observability features to widen the moat.

🎯 Implication

  • For PMs running multi-model AI products: evaluate OpenRouter, Portkey, LiteLLM. The routing layer cuts vendor lock-in and adds observability for free.
  • For execs negotiating AI vendor contracts: a routing layer in your stack changes negotiating leverage. Mention it.
·TechStartups - Venture Capital Roundup May 26 (OpenRouter $113M)
FUNDINGOther

Fireworks Hits $15 Billion

27MAY

Inference-infrastructure company Fireworks AI is in talks to raise at a $15 billion valuation (Bloomberg, May 27). The round has not yet closed.

The inference-routing layer just became investable at unicorn scale. Fireworks isn't building models. It's running other people's models faster and cheaper.

Fireworks competes with Together AI, Modal, and Replicate on hosted inference. The $15B mark would 4x its January 2025 valuation. Customers run Llama, Mistral, DeepSeek, and other open-weight models, plus closed-weight calls via API.

Capital is moving down the stack from models to infrastructure. Watch the next big customer signing. That's the real signal.

full brief & sources

⚡ Why this matters

  • Inference is becoming a layer worth $15B before any one company has clearly won.
  • Open-weight model deployment economics now matter as much as the models themselves.
  • If Fireworks closes at this number, expect Together AI to follow with a comparable round.

🔍 What happened

  • May 27, 2026. Bloomberg reports Fireworks AI in talks for a funding round at $15B valuation.
  • Round has not yet closed as of reporting.
  • Fireworks runs open-weight models (Llama, Mistral, DeepSeek, Qwen) as a hosted inference platform.
  • Competes with Together AI, Modal, Replicate, Anyscale.
  • Previous valuation was around $3.5-4B in January 2025.

💬 Smart takes

  • Bloomberg: 'a startup that helps companies run artificial intelligence models, is in talks to raise a new round of funding'
  • Industry framing: investors are paying up for the inference layer, not just the model layer.
  • Skeptic: $15B is a 4x mark on a company whose moat is operational efficiency, not technology. Competitive pressure from Together AI and the hyperscalers' own offerings is real.

🧭 Where this goes

  1. Round closes within 60 days at the reported valuation or close to it.
  2. Together AI raises a comparable round at $10-12B within 90 days.
  3. Hyperscalers (AWS Bedrock, Azure AI Foundry, GCP Vertex) sharpen their hosted-inference pricing in response.
  4. Open-weight model labs (Mistral, DeepSeek, Alibaba Qwen) deepen partnerships with inference platforms.

🎯 Implication

  • For PMs running AI vendor evaluation: add Fireworks and Together to your bake-off for any open-weight workload. Cheaper than Anthropic or OpenAI API for equivalent capability.
  • For execs tracking AI infrastructure costs: inference layer pricing is becoming competitive. Renegotiate hosted-inference contracts in Q3.
·Bloomberg - Fireworks AI in Talks for Funding at $15 Billion Valuation
Tuesday May 26
FUNDINGAnthropic

Anthropic's First Profitable Quarter

20MAY
FIRST PROFIT!!$559M Q2 - YEP, REAL$10.9BQ2 REVENUE+130% Q-o-Q$559MOPERATINGPROFITANTHROPIC

Anthropic projects $10.9 billion Q2 revenue (up 130% from $4.8B in Q1) and a $559 million operating profit. First profitable quarter in the lab's history.

OpenAI is still burning cash at frontier scale. Anthropic just stopped.

The $559M profit excludes stock-based comp and pre-paid compute commitments. Ed Zitron (Where's Your Ed At) and other skeptics call it a "profitability swindle." Either way, the revenue line is real and the growth rate outpaces historical peaks at Zoom, Google, and Facebook.

For the $900B valuation conversation, this turns the IPO from narrative into cash flow. Enterprise risk on 5-year Claude bets just dropped.

full brief & sources

⚡ Why this matters

  • First positive operating quarter ever from a frontier AI lab. That's the line the market hasn't priced.
  • Revenue growth (130% QoQ) outpaces Zoom, Google, and Facebook at their historical peaks.
  • Turns the $900B valuation conversation from narrative into cash flow.

🔍 What happened

  • May 20, 2026. Anthropic informs investors of Q2 projections (Bloomberg, CNBC, The Information).
  • $10.9B Q2 revenue projected. Up from $4.8B Q1.
  • $559M operating profit projected. First profitable quarter ever.
  • Profit calculation excludes stock-based compensation and pre-paid compute (the way most frontier labs would report).
  • Investor talks underway on a $900B valuation funding round. Above OpenAI's $852B March mark.

💬 Smart takes

  • Bloomberg / CNBC: "Quarterly growth rate currently outpaces historical peaks of Zoom, Google, and Facebook."
  • Ed Zitron (Where's Your Ed At, vocal AI critic): calls the framing a "profitability swindle" because of the excluded compute and equity costs.
  • Skeptic read: Anthropic may not stay profitable across the year. High scheduled compute costs (per the SpaceX S-1) load up in H2 2026.

🧭 Where this goes

  1. Anthropic closes its $900B round by end-May with this data point doing the heavy lifting.
  2. IPO timing accelerates. 2026 H2 or early 2027 looks plausible now, not 2028.
  3. OpenAI faces pressure to publish equivalent profit projections. Sam Altman's "we're losing money on everyone" framing gets stale.
  4. Compute-spend-as-percent-of-revenue (per SpaceX S-1) becomes the diligence question: is the profit margin sustainable when compute commitments fully load?

🎯 Implication

  • For enterprises betting on Claude: Anthropic's runway risk just dropped. The lab will be around to support 5-year contracts.
  • For exec readers tracking AI vendor risk: add "frontier-lab operating profitability" to your vendor stability scorecard. Anthropic is the first to clear this bar.
·CNBC - Anthropic set to hit $10.9 billion in revenue in Q2 (May 20)·Dataconomy - Anthropic projects $10.9B Q2 revenue and first ever profit·Where's Your Ed At - Anthropic's "Profitability" Swindle (skeptic)