NVIDIA, directly or through its venture arm NVentures, joined 62 AI startup funding rounds in 2025, counting each funding announcement once. The 61 rounds with a disclosed size add up to $22.6B. That figure is the full size of each round. NVIDIA’s own share is rarely disclosed.
Three numbers describe the year. NVIDIA was named a lead investor in only two of the 62 rounds, and press coverage called it a co-lead in a third. Nine rounds of $1B or more carried 65% of the combined value. And 49 of the 62 rounds were announced in the second half, 17 of them in September.
On this page
NVIDIA’s 2025 AI deals at a glance
| Metric | 2025 |
|---|---|
| AI startup rounds NVIDIA or NVentures joined | 62 |
| Combined size of the 61 disclosed rounds | $22.6B |
| Median round size | $115M |
| Rounds of $100M or more | 36 |
| Rounds of $1B or more | 9 ($14.7B combined) |
| Rounds naming NVIDIA as a lead | 2 (Reflection AI, Factory), plus 1 reported co-lead |
| Rounds for US-based companies | 46 |
| Busiest month | September (17 rounds) |
Source: Bot Memo. Round sizes are full round totals. Rounds in euros and Australian dollars are converted to US dollars.
The nine billion-dollar rounds were Cursor ($2.3B Series D), Thinking Machines Lab ($2B seed), Reflection AI ($2B Series B), Safe Superintelligence ($2B), Mistral AI (€1.7B Series C), Crusoe ($1.375B Series E), Nscale ($1.1B Series B), Figure (more than $1B Series C) and PsiQuantum ($1B Series E). NVIDIA led Reflection AI’s. In the other eight it was one investor among many.
For how other large tech companies put money into AI startups, see our analysis of corporate VC AI investments.
Every AI round NVIDIA joined in 2025
The full list, largest round first. Amounts are the total round size. “Unlabeled” means the company did not name a stage.
| Company | Round size | Stage | Month | Headquarters | What it does |
|---|---|---|---|---|---|
| Cursor (Anysphere) | $2.3B | Series D | November | San Francisco, United States | AI code editor and coding agents |
| Reflection AI | $2B (NVIDIA led) | Series B | October | New York, United States | Open frontier models and coding agents |
| Safe Superintelligence | $2B | Unlabeled | April | Palo Alto, United States | Research lab building safe superintelligence |
| Thinking Machines Lab | $2B | Seed | July | San Francisco, United States | Frontier AI research lab |
| Mistral AI | €1.7B (about $1.9B) | Series C | September | Paris, France | Open and commercial AI models |
| Crusoe | $1.375B | Series E | October | Denver, United States | AI data centers and cloud |
| Nscale | $1.1B | Series B | September | London, United Kingdom | AI cloud and data centers |
| Figure | $1B+ | Series C | September | San Jose, United States | Humanoid robots |
| PsiQuantum | $1B | Series E | September | Palo Alto, United States | Photonic quantum computers |
| Quantinuum | $600M | Unlabeled | September | Broomfield, CO, United States | Trapped-ion quantum computers |
| Cohere | $500M | Series D | August | Toronto, Canada | Enterprise AI models and agents |
| Lambda | $480M | Series D | February | San Francisco, United States | GPU cloud for AI |
| Nscale | $433M | Pre-Series C SAFE | October | London, United Kingdom | AI cloud and data centers |
| FieldAI | $405M | Unlabeled | August | Irvine, United States | AI models that run field robots |
| Lovable | $330M | Series B | December | Stockholm, Sweden | Builds apps from text prompts |
| Runway | $308M | Series D | April | New York, United States | AI video generation |
| Together AI | $305M | Series B | February | San Francisco, United States | Cloud for open-source AI models |
| Black Forest Labs | $300M | Series B | December | Freiburg, Germany | Image generation models (FLUX) |
| Periodic Labs | $300M | Seed | September | San Francisco, United States | Autonomous labs for science |
| Uniphore | $260M | Series F | October | Palo Alto, United States | Enterprise AI agent platform |
| Abridge | $250M | Series D | February | Pittsburgh, United States | AI medical scribe |
| Suno | $250M | Series C | November | Cambridge, MA, United States | AI music generation |
| Firmus | A$330M (about $215M) | Unlabeled | September | Singapore, Singapore | AI data centers in Tasmania |
| n8n | $180M | Series C | October | Berlin, Germany | Workflow automation and AI agents |
| Aidoc | $150M (incl. about $40M credit line) | Unlabeled | July | New York, United States | Clinical AI for medical imaging |
| SandboxAQ | $150M (extension; Series E total over $450M) | Series E extension | April | Palo Alto, United States | Quantitative AI models for science and security |
| Hippocratic AI | $141M | Series B | January | Palo Alto, United States | AI agents for patient-facing tasks |
| fal | $140M | Series D | December | San Francisco, United States | Infrastructure for generative media models |
| Reflection AI | $130M | Seed and Series A | March | New York, United States | Open frontier models and coding agents |
| Dyna Robotics | $120M | Series A | September | Redwood City, United States | Robot foundation models |
| Lila Sciences | $115M | Series A extension | October | Cambridge, MA, United States | AI-run autonomous science labs |
| Rescale | $115M | Series D | April | San Francisco, United States | Cloud for engineering simulation |
| Reka | $110M | Unlabeled | July | United States | Multimodal AI models |
| Cartesia | $100M | Unlabeled | October | San Francisco, United States | Real-time voice AI models |
| CuspAI | $100M | Series A | September | Cambridge, United Kingdom | AI search for new materials |
| Perplexity | $100M | Unlabeled | July | San Francisco, United States | AI answer engine |
| Nuro | $97M (August tranche; Series E total $203M) | Series E | August | Mountain View, United States | Self-driving software |
| PolyAI | $86M | Series D | December | London, United Kingdom | Voice AI agents for customer service |
| Exa | $85M | Series B | September | San Francisco, United States | Search engine for AI agents |
| Adaptive Security | $81M | Series B | December | New York, United States | Defense against deepfake attacks |
| Bedrock Robotics | $80M | Seed and Series A | July | San Francisco, United States | Self-driving kits for construction machines |
| Charm Therapeutics | $80M | Series B | September | London, United Kingdom | AI-designed cancer drugs |
| PassiveLogic | $74M | Series C | September | Salt Lake City, United States | Autonomous control of building systems |
| Luminary Cloud | $72M | Series B | September | San Mateo, United States | Physics AI for engineering simulation |
| Utilidata | $60.3M | Series C | April | Providence, United States | AI for smart grid meters |
| CodeRabbit | $60M | Series B | September | San Francisco, United States | AI code review |
| Scintil Photonics | $58M | Series B | September | Grenoble, France | Photonic chips for AI data centers |
| Radical AI | $55M | Seed | July | New York, United States | Autonomous labs for new materials |
| Factory | $50M (NVIDIA named among leads) | Series B | September | San Francisco, United States | AI software-engineering agents |
| Flexion | $50M | Series A | November | Zurich, Switzerland | Software for humanoid robots |
| Inception | $50M | Seed | November | Palo Alto, United States | Diffusion language models |
| Phaidra | $50M+ | Series B | October | Seattle, United States | AI control of data-center cooling |
| WisdomAI | $50M | Series A | November | San Mateo, United States | AI analytics on company data |
| Achira | $33M | Seed | February | Not stated | AI and physics models for drug discovery |
| Moonlake AI | $28M (reported as co-led by NVIDIA) | Seed | October | San Francisco, United States | AI tools to build game worlds |
| Relation Therapeutics | $26M | Unlabeled | December | London, United Kingdom | AI drug discovery |
| Emerald AI | $24.5M | Seed | July | United States | Flexible data-center power use |
| Simular | $21.5M | Series A | December | United States | AI agents that operate a computer |
| PhysicsX | $20M | Series B extension | November | London, United Kingdom | AI simulation for engineering |
| tryprofound.com | $20M | Series A | June | New York, United States | Tracks brands in AI search answers |
| GamerBoom | $9M | Unlabeled | February | New York, United States | AI gaming data analytics |
| QuEra Computing | Undisclosed | Series B extension | September | Boston, United States | Neutral-atom quantum computers |
Source: Bot Memo. QuEra’s extension amount was not disclosed; it expanded a $230M Series B first announced in February 2025.
The rounds NVIDIA led
Reflection AI is the largest. NVIDIA led its $2B Series B in October 2025 (some reports name Lightspeed as a co-lead), which valued the New York company at $8B. Reflection builds open frontier models and coding agents. NVentures had also joined the company’s $130M seed and Series A in March.
The other two were small. Factory, which builds AI agents for software teams, listed NVIDIA among the investors that led its $50M Series B in September, alongside NEA, Sequoia Capital and J.P. Morgan. Moonlake AI, which makes tools for building game worlds, raised a $28M seed in October from AIX Ventures, Threshold and NVentures; press coverage described the three as co-leads.
In the other 59 rounds, no announcement names NVIDIA as a lead.
Where the money went
Companies building AI models and developer tools took $11.7B across 15 rounds, just over half of the combined value. Four of the nine billion-dollar rounds were model labs: Thinking Machines Lab, Reflection AI, Safe Superintelligence and Mistral AI. Coding tools add Cursor, Lovable ($330M), CodeRabbit ($60M) and Factory.
Compute and data-center companies raised $4.3B across 10 rounds. Crusoe, Nscale (twice), Lambda ($480M Series D), Together AI ($305M) and Firmus (A$330M) build or rent out GPU capacity. Companies like these buy large amounts of AI hardware, so their growth can also mean more chip sales for NVIDIA. Critics call stakes like these circular for that reason.
Science, health and biology had the second-most rounds, 12, but only $1.4B combined. The largest were Periodic Labs ($300M seed), Abridge ($250M Series D), SandboxAQ ($150M), Aidoc ($150M) and Hippocratic AI ($141M).
Robotics and physical AI drew $1.8B across 7 rounds: Figure, FieldAI ($405M), Nuro ($97M), Dyna Robotics ($120M), Bedrock Robotics ($80M), PassiveLogic ($74M) and Flexion ($50M). Quantum computing, chips and power added $1.7B across 6, led by PsiQuantum and Quantinuum ($600M).
For the wider market these deals sit in, see our 2025 AI funding year in review.
Where the companies are
46 of the 62 rounds went to US-based companies, worth $17.2B combined. Europe had 13, led by the UK with 7.
| Country | Rounds | Companies |
|---|---|---|
| United States | 46 | Cursor, Thinking Machines Lab, Reflection AI (2 rounds), Safe Superintelligence, Crusoe and 40 other companies |
| United Kingdom | 7 | Nscale (2 rounds), CuspAI, PolyAI, Charm Therapeutics, Relation Therapeutics, PhysicsX |
| France | 2 | Mistral AI, Scintil Photonics |
| Germany | 2 | Black Forest Labs, n8n |
| Canada | 1 | Cohere |
| Sweden | 1 | Lovable |
| Switzerland | 1 | Flexion |
| Singapore | 1 | Firmus |
Source: Bot Memo. Not shown: Achira, a drug-discovery startup with no stated headquarters.
All seven UK rounds came from September onward. On September 18, 2025 NVIDIA announced a £2B commitment to back UK AI startups, alongside Accel, Air Street Capital, Balderton, Hoxton Ventures and Phoenix Court. Nscale’s $1.1B Series B followed a week later, and its $433M pre-Series C SAFE (a convertible instrument rather than a priced round) came on October 1. CuspAI ($100M, materials discovery), PolyAI ($86M, voice agents for customer service), Charm Therapeutics ($80M, cancer drugs), Relation Therapeutics ($26M) and PhysicsX ($20M, engineering simulation) make up the rest.
To see how a different corporate investor split its European bets, compare every AI deal Google Ventures made in 2025.
Every stage, from seed to Series F
| Stage | Rounds |
|---|---|
| Seed | 7 |
| Seed and Series A together | 2 |
| Series A (including 1 extension) | 7 |
| Series B (including 2 extensions) | 16 |
| Series C (including 1 pre-Series C SAFE) | 7 |
| Series D | 8 |
| Series E and F (including 1 extension) | 5 |
| Unlabeled | 10 |
Source: Bot Memo.
Series B was the most common stage, with 16 rounds. Thinking Machines Lab’s $2B seed is bigger than every Series D on this list except Cursor’s, and Periodic Labs raised $300M at seed. The smaller seeds were Radical AI ($55M, materials labs), Inception ($50M, diffusion language models), Achira ($33M), Moonlake AI ($28M) and Emerald AI ($24.5M, flexible data-center power use).
For a portfolio built differently, see Microsoft’s 2025 AI investments.
What NVIDIA announced but did not close in 2025
NVIDIA’s largest 2025 headlines were commitments that had not closed by year end, so they are not in the count above.
- OpenAI. On September 22, 2025 NVIDIA signed a letter of intent to invest up to $100B in OpenAI, in stages as each gigawatt of NVIDIA systems is deployed. No tranche closed in 2025.
- Anthropic. On November 18, 2025 NVIDIA committed to invest up to $10B, with Microsoft committing up to $5B. Both were commitments; no money from them was counted as a 2025 round.
- xAI. NVIDIA joined xAI’s $20B Series E, which closed in January 2026.
FAQ
Which AI startups did NVIDIA invest in during 2025?
NVIDIA or NVentures joined 62 AI startup rounds in 2025. The largest were Cursor ($2.3B), Thinking Machines Lab ($2B), Reflection AI ($2B), Safe Superintelligence ($2B), Mistral AI (€1.7B), Crusoe ($1.375B), Nscale ($1.1B), Figure (more than $1B) and PsiQuantum ($1B). The full list is in the table above. For other active backers, see the most active AI investors in 2025.
How much did NVIDIA invest in AI startups in 2025?
NVIDIA participated in rounds worth $22.6B in total, counting the 61 with a disclosed size. Its own share of that is mostly undisclosed.
Did NVIDIA lead most of its AI deals?
No. It led Reflection AI’s $2B Series B, and Factory named it among the lead investors of its $50M Series B. Press coverage described it as a co-lead of Moonlake AI’s $28M seed. In the other 59 rounds, no announcement names NVIDIA as a lead.
What kinds of AI companies does NVIDIA back?
Mostly companies that use large amounts of compute. Model labs and developer tools took just over half of the round value, and compute and data-center companies another $4.3B. Robotics, quantum computing, drug discovery and health AI make up most of the rest.
Methodology
What counts. Each funding announcement counts as one round. It must be made between January 1 and December 31, 2025 by an AI startup, and must name NVIDIA or NVentures as an investor in that specific round. A company listing NVIDIA among earlier backers does not count. Where one announcement covered two rounds, it counts once: Reflection AI and Bedrock Robotics announced a seed and Series A together, and FieldAI announced $405M across two back-to-back rounds.
What is left out. Commitments and letters of intent that did not close in 2025 (OpenAI, Anthropic), rounds that closed in 2026 (xAI), rounds whose size could not be confirmed (Skild AI, which was reported in June 2025 to have raised a Series B with NVIDIA but never announced it), share purchases from existing holders, and companies outside AI, such as the fusion, nuclear and battery companies NVIDIA backed in 2025. Quantum computing companies are included as compute businesses.
Amounts. Each figure is the round size as announced. Where an extension added money to a round that opened before 2025, only the 2025 extension counts (SandboxAQ). Where NVIDIA joined only a later tranche or extension of a 2025 round, that tranche counts (Lila Sciences, Nuro, PhysicsX, QuEra). Aidoc’s $150M includes about $40M of credit. Amounts given as “more than” count at the stated floor. QuEra’s extension has no disclosed size and is left out of dollar totals.
Sectors. Each round is placed in one group based on what the company builds. Group totals add up to $22.6B.
About the author
Editorial Staff
The Editorial Staff at Bot Memo is a team of writers, analysts, and AI agents dedicated to mapping the global AI startup ecosystem. Led by Chintan Zalani, the team tracks thousands of funding rounds, classifies companies across verticals, and distills it all into actionable intelligence for investors and founders.
