World
Start with three numbers. In the third quarter of 2026, startups worldwide received $159 billion of venture money, and AI startups took $102 billion of it, or 64%, according to Crunchbase.
If you need one number for the board, take the share: 64% of world venture money in the third quarter went to AI startups. It needs no explanation of currency or of which database counted the deals.
Half-year numbers come from other sources. KPMG counted $560.3 billion of total venture money in the first half of 2026, and PitchBook $407 billion for AI startups in the same period. Compare figures from different databases with care: their counting rules differ.
For the scale of all AI spending, not only venture money, see the Gartner forecast: $2.59 trillion in 2026. This is buyer spending, not investor money, so it is never added to venture numbers.
The Stanford AI Index gives the total for 2025: $581.69 billion of corporate investment in AI. How spending will change next is covered on the forecast page.
Global venture investment in Q3 2026: $159 billion across about 6,000 funded startups; $679 billion for Q1–Q3; $212 billion in Q2; $104 billion in Q3 2025.
Crunchbase News, Gené Teare, 5 October 2026, checked 14 October 2026
AI startups received $102 billion in Q3 2026, or 64% of global venture investment.
Crunchbase News, 5 October 2026, checked 14 October 2026
Rounds of $1 billion or more in Q3 2026: 27 companies, a record, against 16 in Q2 and 14 in Q1. These 27 companies took about a third of the quarter's global venture money.
Crunchbase News, 5 October 2026, checked 14 October 2026
The US in Q3 2026: $91 billion, about 57% of global venture investment; the San Francisco area alone: 24%.
Crunchbase News, 5 October 2026, checked 14 October 2026
Late-stage funding in Q3 2026: $105 billion; early stage: $40.6 billion (+25% year on year); seed: $13 billion.
Crunchbase News, 5 October 2026, checked 14 October 2026
AI startups raised $407 billion in the first half of 2026, against $264.1 billion in all of 2025; about $217 billion of it went to OpenAI and Anthropic.
Kommersant, citing PitchBook, Venera Petrova, 13 August 2026, checked 14 October 2026
Global venture investment in the first half of 2026: $560.3 billion, 2.2 times more than a year earlier (Q1 $332.9 billion, Q2 $227.4 billion).
Kommersant, citing KPMG Private Enterprise, 13 August 2026, checked 14 October 2026
Global corporate investment in AI in 2025: $581.69 billion, a record; private investment $344.66 billion; mergers and acquisitions $214.44 billion. US $285.88 billion, Europe $20.92 billion, China $12.41 billion.
Stanford HAI, AI Index 2026, chapter 'Economy', 13 April 2026, checked 14 October 2026
Global AI spending in 2026 as forecast by Gartner: $2.59 trillion (+47%), after $1.76 trillion in 2025 and before $3.49 trillion in 2027. AI infrastructure in 2026: $1.43 trillion.
Gartner, press release, 19 May 2026, checked 14 October 2026
McKinsey survey of 1,719 participants in 97 countries (May–June 2026): 44% of organizations are scaling AI across the whole company; 37% see at least some AI contribution to EBIT; about 6% attribute at least 5% of EBIT to AI; 60% expect AI investment to grow next year; 28% spend more than 10% of their ICT budget on AI; 32% dropped the purchase of at least one product because they can build it themselves with agents; 40% of large companies are scaling AI agents.
McKinsey, The state of AI in 2026, August 2026, checked 14 October 2026
A report by Russia's Roscongress Foundation put the chance of a crash in the market for AI companies in 2026 above 50%; since the start of 2024 the S&P 500 gained about $7.5 trillion, $4.9 trillion of it from 17 AI-related companies. It also describes circular financing: chip makers invest in customers that buy their hardware.
Mail.ru Finance, citing RIA Novosti and the Roscongress report, 19 February 2026, checked 14 October 2026
On 21 January 2025 OpenAI, SoftBank, and Oracle announced the Stargate project: $100 billion at once and up to $500 billion over four years for AI infrastructure in the US. This is private money, not the federal budget.
TechCrunch, Kyle Wiggers, 21 January 2025, checked 16 October 2026
On 11 February 2025 the European Commission launched InvestAI, which aims to mobilize €200 billion of investment in AI, including a new €20 billion European fund for AI gigafactories.
European Commission, Digital Strategy, 11 February 2025, checked 16 October 2026
On 30 July 2026 EuroHPC opened a call to build up to seven AI gigafactories in the EU. Expected private investment: more than €20 billion; applications are accepted until 12 November 2026.
EuroHPC Joint Undertaking, 30 July 2026, checked 16 October 2026
In February 2025 Emmanuel Macron announced €109 billion of private AI investment in France. The largest commitments came from the UAE, with €30–50 billion, and Canada's Brookfield, with €20 billion.
TechCrunch, Romain Dillet, 10 February 2025, checked 16 October 2026
The UK government has allocated more than £2 billion to computing infrastructure for AI: over £1 billion to expand the AI Research Resource at least 20-fold by 2030, and up to £750 million for a national supercomputer in Edinburgh. Figures from an announcement of 26 January 2026.
GOV.UK, Department for Science, Innovation and Technology, 26 January 2026, checked 16 October 2026
China has launched a state investment fund for the AI industry worth 60 billion yuan. The Xinhua news agency reported the launch on 21 January 2026.
Xinhua, 21 January 2026, checked 16 October 2026
On 26 July 2026 Russia's president signed a law on the development of AI technologies. It governs how large foundation models, meaning models with at least 1 billion parameters, are developed, deployed, and used.
Interfax, 26 July 2026, checked 16 October 2026
11 September 2026: an investment tax deduction for building AI models is under discussion, up to $240 million per organization until 2030. The measure has not been adopted.
Vedomosti, 11 September 2026, checked 16 October 2026
The largest companies
The main number of the section: Amazon, Microsoft, Alphabet and Meta plan capital spending of $700–725 billion in 2026, against $410 billion in 2025, according to a Forbes estimate.
An industry review gives a similar picture: the capital spending of hyperscalers, the largest data center owners, grew by about 80% in a year, and more than $1 trillion is expected next year.
How to read these numbers: capital spending means data centers and equipment. It is not investment in startups and not AI revenue, so it is not compared with round sizes.
A breakdown for each company, with quarters and plans, is in the article on the capital spending of the giants.
Capital spending of the four largest companies in 2026: Amazon about $200 billion, Microsoft about $190 billion, Alphabet $180–190 billion, Meta $125–145 billion; together $700–725 billion, against $410 billion in 2025 (about +75%).
Forbes, Jason Kirsch, 7 October 2026, checked 14 October 2026
Capital spending of the hyperscalers, the largest cloud providers, grew about 80% in a year; more than $1 trillion is expected next year.
Yandex Cloud blog, webinar 'A look at the AI market', 24 July 2026, checked 14 October 2026
In fiscal year 2026 (July 2025 to June 2026) Microsoft spent $115.9 billion on equipment and construction, against $64.6 billion a year earlier; $35.8 billion of it fell in April–June 2026.
Microsoft, press release on Q4 fiscal 2026 results, 29 July 2026, checked 16 October 2026
On 26 January 2026 Microsoft unveiled Maia 200, its own AI accelerator for running trained models: built on TSMC's 3 nm process, it gives 30% more performance per dollar than the company's other newest hardware. The US Central data center region in Iowa got it first.
Microsoft, official blog, Scott Guthrie, 26 January 2026, checked 16 October 2026
After OpenAI's restructuring on 28 October 2025, Microsoft owns about 27% of OpenAI Group PBC, a stake valued at about $135 billion. In return, OpenAI committed to buy an extra $250 billion of Azure cloud services.
Microsoft, official blog, 28 October 2025, checked 16 October 2026
On 15 April 2026 Microsoft agreed to rent another 30,000 Nvidia Vera Rubin chips from Nscale in Narvik, Norway, at a site first prepared for OpenAI's Stargate project. Microsoft had already taken on $6.2 billion of commitments for this site.
Bloomberg News, Brody Ford, Kurt Wagner, Mark Bergen, 15 April 2026, checked 16 October 2026
In Q2 2026 Alphabet spent $44.9 billion on equipment and buildings, and $80.6 billion in the first half of the year.
Alphabet, press release on Q2 2026 results (SEC Form 8-K), 22 July 2026, checked 16 October 2026
On 24 April 2026 Alphabet said it would invest up to $40 billion in Anthropic: $10 billion in cash at once at a $350 billion valuation, and another $30 billion if Anthropic meets its targets.
Reuters (Deborah Sophia, Zaheer Kachwala), republished by CP24, 24 April 2026, checked 16 October 2026
In June 2026 Alphabet sold shares and convertible preferred shares for $49.6 billion net. Among the purposes, the company named capital spending on AI infrastructure and computing capacity.
Alphabet, press release on Q2 2026 results (SEC Form 8-K), 22 July 2026, checked 16 October 2026
In Q2 2026 Amazon spent $54.2 billion on equipment and buildings. AWS revenue grew 37%, to $42.2 billion.
Amazon, press release on Q2 2026 results, 30 July 2026, checked 16 October 2026
In Q2 2026 Amazon booked $53.4 billion of other income before tax, mostly from revaluing its stake in Anthropic.
Amazon, press release on Q2 2026 results, 30 July 2026, checked 16 October 2026
In Q2 2026 Meta's capital spending was $31.08 billion, and free cash flow fell to $784 million. Meta narrowed its 2026 capital spending forecast to $130–145 billion from $125–145 billion.
Meta, press release on Q2 2026 results, 29 July 2026, checked 16 October 2026
On 12 June 2025 Meta invested in Scale AI, a company that labels data for model training, at a valuation of more than $29 billion for Scale. After the deal, Meta holds a minority stake in Scale.
Scale AI, official blog, 12 June 2025, checked 16 October 2026
On 24 February 2026 Meta and AMD agreed to deploy up to 6 GW of AMD Instinct accelerators, with first deliveries in the second half of 2026. AMD gave Meta a warrant to buy up to 160 million AMD shares, which vests as the deal reaches its milestones.
AMD, press release, 24 February 2026, checked 16 October 2026
On 17 February 2026 Meta and Nvidia signed a multiyear agreement covering several generations of hardware: millions of Nvidia Blackwell and Rubin GPUs for Meta's new large data centers, built both for training and for running models.
Nvidia, press release, 17 February 2026, checked 16 October 2026
In Q2 of fiscal 2027 (May–July 2026) Nvidia bought $15.8 billion of shares in other companies, and $42.4 billion over the half year. Non-public stakes on its balance sheet grew to $51.2 billion from $22.3 billion in January 2026.
Nvidia, press release on Q2 fiscal 2027 results, 26 August 2026, checked 16 October 2026
Nvidia's data center revenue in Q2 of fiscal 2027 was $89.0 billion, 117% more than a year earlier. The Vera Rubin platform, according to Jensen Huang, is in full production.
Nvidia, press release on Q2 fiscal 2027 results, 26 August 2026, checked 16 October 2026
On 29 December 2025 it became known that Nvidia had bought $5 billion of Intel shares: more than 214.7 million shares at $23.28 each.
Reuters, republished by Cyprus Mail, 29 December 2025, checked 16 October 2026
On 26 January 2026 Nvidia invested another $2 billion in the cloud provider CoreWeave at $87.20 a share, almost doubling its stake, which was 6.3% before the deal. CoreWeave is building more than 5 GW of data center capacity by 2030.
Reuters (Deborah Sophia), republished by CP24, 26 January 2026, checked 16 October 2026
In its report of 26 August 2026 Nvidia described compute financing platforms with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR: over time they are to raise more than $500 billion of outside capital for AI infrastructure.
Nvidia, press release on Q2 fiscal 2027 results, 26 August 2026, checked 16 October 2026
On 22 April 2026 Google unveiled the eighth generation of its TPU AI chips. One TPU 8t supercluster links 9,600 chips and delivers 121 exaflops, while the TPU 8i for running trained models gives 80% more performance per dollar than the previous generation, by Google's figures. Both chips will be generally available by the end of 2026.
Google, official blog, Amin Vahdat, 22 April 2026, checked 16 October 2026
According to Google's technical description, the new Virgo network lets a single training cluster hold more than 1 million TPU chips.
Google Cloud, official blog, Diwakar Gupta, Sabastian Mugazambi, 22 April 2026, checked 16 October 2026
On 22 July 2026 Alphabet raised its 2026 capital spending forecast to $195–205 billion. According to Bloomberg, the previous forecast was up to $190 billion.
Bloomberg News, Davey Alba, Julia Love (Bloomberg Law), 23 July 2026, checked 16 October 2026
Anthropic trains and runs Claude on more than 1 million Amazon Trainium2 chips. Under an agreement of 20 April 2026, almost 1 GW of Trainium2 and Trainium3 capacity is to come online by the end of 2026, and up to 5 GW over time.
Anthropic, official website (news), 20 April 2026, checked 16 October 2026
On 20 April 2026 Amazon invested $5 billion in Anthropic and may add up to $20 billion more. Anthropic committed to spend more than $100 billion on AWS technology over 10 years.
Anthropic, official website (news), 20 April 2026, checked 16 October 2026
On 13 July 2026 Meta said it would expand its data center campus in Richland Parish, Louisiana, to 5 GW of computing capacity. Its investment in the region will exceed $50 billion, with more than $1 billion more for local roads and water supply.
Meta, official website (news), 13 July 2026, checked 16 October 2026
On 28 July 2026 Meta announced a 1 GW data center in El Paso, Texas, with Meta's investment above $10 billion. A joint venture builds it, with BlackRock funds owning 80% and Meta 20%.
Meta, official website (news), 28 July 2026, checked 16 October 2026
On 18 November 2025 Microsoft pledged to invest up to $5 billion in Anthropic. Anthropic committed to buy $30 billion of Azure computing capacity and to contract up to 1 GW more.
Anthropic, official website (news), 18 November 2025, checked 16 October 2026
On 18 November 2025 Nvidia pledged to invest up to $10 billion in Anthropic. The companies agreed on joint engineering work, including tuning future Nvidia architectures for Anthropic's workloads.
Anthropic, official website (news), 18 November 2025, checked 16 October 2026
On 30 July 2026 Amazon raised its 2026 capital spending forecast to $220 billion from $200 billion, citing more expensive memory. Andy Jassy said that even this spending will not give enough capacity for all of 2026 demand.
Fortune, Amanda Gerut, 30 July 2026, checked 16 October 2026
Deals
Two deals set the scale of the year. OpenAI closed a $122 billion round at a valuation of $852 billion, and Anthropic raised $65 billion at a valuation of $965 billion.
Among third-quarter acquisitions, the largest by amount is the Nvidia offer to buy Hugging Face for $12.9 billion. The source reports an offer, not a closed deal, and the list below shows it the same way.
How to read these numbers: a company’s valuation and the size of a round are different things. The valuation shows what investors think the whole company is worth; the round shows how much money it received.
Single deals, with filters by region and type, are in the deal tracker.
OpenAI closed a $122 billion round at an $852 billion valuation. Lead investors: Amazon $50 billion (about $35 billion of it depends on an IPO or a technology milestone), Nvidia $30 billion, SoftBank $30 billion.
Bloomberg Law, 1 April 2026, checked 14 October 2026
Anthropic raised $65 billion at a $965 billion valuation and passed OpenAI in valuation for the first time. Lead investors: Altimeter Capital, Dragoneer, Greenoaks, Sequoia Capital; Google put in several billion dollars under an earlier commitment.
Bloomberg Law, 28 May 2026, checked 14 October 2026
The largest rounds of Q3 2026: Databricks $5 billion, Safe Superintelligence $5 billion; Crusoe, Moonshot AI, Mistral AI, Nscale, and Kling AI each raised more than $3 billion.
Crunchbase News, 5 October 2026, checked 14 October 2026
The largest deals of Q2 2026 according to KPMG: Anthropic $65 billion, Project Prometheus $12 billion, DeepSeek $7.4 billion; in defense AI, Anduril $5 billion.
Kommersant, citing KPMG, 13 August 2026, checked 14 October 2026
Mergers and acquisitions in Q3 2026: Nvidia's offer to buy Hugging Face for $12.9 billion; AMD's plan to buy World Labs for $8.2 billion; Stripe's purchase of OpenRouter for $7.5 billion.
Crunchbase News, 5 October 2026, checked 14 October 2026
IPOs of Q3 2026: ChangXin Memory Technologies raised $8.6 billion at an $85.5 billion valuation; Enflame $912 million at $25.5 billion; Unitree Robotics $905 million at $9 billion; Bending Spoons $1.6 billion at $18.4 billion.
Crunchbase News, 5 October 2026, checked 14 October 2026
Deals with AI companies in 2026 per Solganick: Autodesk bought MaintainX for $3.6 billion (May), Roche bought PathAI for more than $750 million (May), Anthropic bought Stainless for more than $300 million (May); about 305 AI deals worldwide in Q1 2026 (+90% year on year).
Solganick, AI M&A update YTD 2026, June 2026, checked 14 October 2026
Project Prometheus was valued at $40 billion as of July 2026; Anthropic bought about one million TPUs from Google Cloud.
Yandex Cloud blog, 24 July 2026, checked 14 October 2026
CuspAI (AI for materials science, UK) raised $450 million in a Series B at a $2.6 billion valuation in July 2026; investors include Kleiner Perkins, NEA, and Bezos Expeditions.
vc.ru, 2026, checked 14 October 2026
According to Crunchbase, as of 22 June 2026 robotics startups had raised $18.8 billion since the start of the year. That is more than in all of 2025 ($15 billion) and in the peak year of 2021 ($14.1 billion).
Crunchbase News, Mary Ann Azevedo, 22 June 2026, checked 16 October 2026
On 11 February 2026 humanoid robot developer Apptronik raised $520 million in a Series A-X round. The full Series A exceeded $935 million; new investors include John Deere, AT&T Ventures, and Qatar Investment Authority.
Apptronik, press release, 11 February 2026, checked 16 October 2026
In Q3 2026 robotics was one of five sectors that each raised at least $10 billion of venture money in the quarter, according to Crunchbase.
Crunchbase News, Gené Teare, 5 October 2026, checked 16 October 2026
On 11 March 2026 Google closed its purchase of Wiz, a cloud security developer, for $32 billion in cash, a year after the deal was announced.
TechCrunch, Rebecca Bellan; date confirmed by an Alphabet press release of 11 March 2026, 11 March 2026, checked 16 October 2026
A Russian developer of AI services for medicine plans to raise $1.8 million from qualified investors, with BITL group as organizer and shares placed on the Zorko platform. The round was announced on 29 September 2026 and had not closed at the date of publication.
ComNews, 29 September 2026, checked 16 October 2026
Russia
The Russian venture market totaled $68 million in 73 deals in the first 9 months of 2026, of which AI and machine learning took $32 million in 33 deals, according to Venture Guide. The data is preliminary, as of 4 October 2026.
Corporate money in Russian AI is many times larger than venture money: Kama Flow estimates the ratio at about 52:1 for 2025. Experts say the method needs checking, so we present this figure as an order of magnitude.
To compare niches inside Russia, see the Venture Guide lines for medicine, corporate programs, manufacturing and robotics: they are in the list below.
How to read these numbers: Russian amounts are shown in US dollars, converted at the Bank of Russia rate of 82.98 per dollar on 13 November 2026 and rounded, and we do not compare Russian venture money with the world market directly. We review deals, funds and state support in the section on Russia.
Russia's venture market in the first half of 2026: $55 million, down about 48% year on year; 54 deals; 33 investors; the largest deal, almost $13 million, went to ArchiTech AI; private funds 67% ($37 million), business angels 16%, corporations 9%, state funds 8%.
Kommersant, citing the Moscow Venture Fund, 13 August 2026, checked 14 October 2026
AI and machine learning in Russia in the first nine months of 2026: $32 million of venture investment in 33 deals, first place by both volume and deal count; ArchiTech AI $13 million (Series C+, investor UKM Group). Preliminary data as of 4 October 2026.
ComNews, citing Venture Guide (Moscow Venture Fund), 5 October 2026, checked 14 October 2026
Kama Flow: by public figures, Russian IT corporations put $2.7 billion into AI in 2025, while venture investment in AI companies was about $51 million (32% of a $159 million venture market); corporate money outweighed venture money about 52 to 1. In 2021 the venture market was $2.5 billion, with $361 million of it in AI. Experts say the method needs checking.
ComNews, citing a Kama Flow study, 6 October 2026, checked 14 October 2026
Kama Flow: an independent AI market in Russia needs about $1.0 billion of private venture money a year; the reachable segments could exceed $6.8 billion by 2030: industrial AI $1.1 billion, AI agents and process automation $940 million, medtech $840 million, finance $480 million, sales and support $480 million, implementation and training $420 million.
ComNews, citing Kama Flow, 6 October 2026, checked 14 October 2026
A 2025 industry study: AI could add $95–150 billion a year to Russia's economy by 2030, up to 5.5% of forecast GDP; the companies with the largest effect spend 13–17% of their IT budget on AI and get 6.6% of EBITDA from it now, with 15.8% expected. Survey of 150 CTOs from 16 industries.
Expert, 8 December 2025, checked 14 October 2026
Russia's national project 'Data Economy and Digital Transformation of the State': $1.9 billion in 2026, more than $2.0 billion in 2027, more than $2.2 billion in 2028 (explanatory note to the draft federal budget).
Mail.ru Finance, citing TASS, 29 September 2025, checked 14 October 2026
Large Russian AI developers, according to RBC, asked the state for about $4.8–5.4 billion a year to develop AI and offered to give 5% of their computing capacity each to education and to government bodies; no decision has been made.
Pravo.ru, citing RBC, 13 March 2026, checked 11 November 2026
Russian AI startups in 2026: Mymeet.ai raised $200,000 from a private investor (late June), Health Hero $180,000; ArchiTech AI (Kodik) $15 million for 20% from UKM Group (June).
vc.ru, 2026, checked 14 October 2026
Russia's market for corporate AI solutions is estimated at $300 million, rising by 2030 to more than $1.1 billion in the base case and up to $2.2 billion in the optimistic one; 88% of companies use AI, but under 10% see an effect above 5% of profit.
Yandex Cloud blog, citing Yakov and Partners, 24 July 2026, checked 14 October 2026
Robotics in Russia in the first nine months of 2026: $13 million of venture investment in 9 deals. Preliminary data as of 4 October 2026.
ComNews, citing Venture Guide (Moscow Venture Fund), Anna Shvetsova, 5 October 2026, checked 16 October 2026
Russia's venture market in the first nine months of 2026: $68 million, 73 deals, an average deal of $1.0 million. Preliminary data as of 4 October 2026.
ComNews, citing Venture Guide (Moscow Venture Fund), Anna Shvetsova, 5 October 2026, checked 16 October 2026
The top niches of Russia's venture market in the first nine months of 2026: healthcare $20 million in 16 deals, enterprise software $19 million in 20 deals, industrial technology about $19 million in 12 deals.
ComNews, citing Venture Guide (Moscow Venture Fund), Anna Shvetsova, 5 October 2026, checked 16 October 2026
Russia's Federal Law No. 243-FZ of 26 July 2026 'On supporting the development of artificial intelligence technologies in the Russian Federation' takes effect on 1 September 2026. Its rules on sovereign and national models and on developers' duties (parts of Articles 5 and 6, Articles 8–10) apply from 1 March 2027.
ConsultantPlus, text of the law, Article 13, 26 July 2026, checked 16 October 2026
Under Law No. 243-FZ, the Russian government will be able to allow only sovereign or national models in certain information systems. For systems already running, this limit does not apply until 1 September 2032 if their data is processed and stored in Russia.
ConsultantPlus, text of the law, Article 13, 26 July 2026, checked 16 October 2026
11 September 2026: a working group gave preliminary approval to state compensation for discounts that small and medium-sized businesses get on Russian AI products. The state could cover up to 100% of the price of a product or annual subscription in the first year, up to 50% in the second, and up to 25% in the third, capped at $2,400 a year, with one client supported for three years. The measure has not been adopted.
Vedomosti, 11 September 2026, checked 16 October 2026
The effect for business
Investment does not equal return. According to the McKinsey 2026 survey, 37% of organizations see at least some contribution of AI to operating profit (EBIT), and about 6% attribute 5% of EBIT or more to AI.
The MIT NANDA report for 2025 gives a harsher number: 95% of organizations got no measurable return from generative AI.
For Russia, an industry study from 2025 estimates the effect of AI on the economy by 2030 at $95–$150 billion a year. This is potential, not today’s result.
How to read these numbers: surveys show that almost everyone uses AI, but only a minority makes money from it. The AI solutions map shows where a return may appear in your company.
MIT NANDA, report The GenAI Divide: State of AI in Business 2025: 95% of organizations got no measurable return from generative AI; only 5% of custom AI tools reach production use.
MIT Media Lab, Project NANDA: The GenAI Divide, State of AI in Business 2025, July 2025, checked 13 October 2026
MIT NANDA 2025: projects with an outside partner reached production use in about 67% of cases, in-house builds in 33%. These are survey answers, not an exact measurement.
MIT Media Lab, Project NANDA: The GenAI Divide, State of AI in Business 2025, July 2025, checked 13 October 2026
An industry survey of 700 Russian companies, December 2025: 91% work with a single AI vendor.
ComNews, citing MTS Web Services, 22 December 2025, checked 13 October 2026
Stanford AI Index 2026: 88% of organizations use AI in at least one business function.
Stanford HAI, The 2026 AI Index Report, 13 April 2026, checked 13 October 2026
A 2025 industry study, a survey of 150 CTOs at large Russian companies from 16 industries: 78% use ready-made AI solutions from outside partners, and 71% fine-tune existing models instead of training their own from scratch.
Yandex and Yakov and Partners: a study of AI in Russia, 8 December 2025, checked 13 October 2026
In January 2026 OpenEvidence, an AI reference service for doctors, raised $250 million at a $12 billion valuation, double the previous one. The company reported annual revenue of $100 million and said more than 40% of US doctors use the service.
Crunchbase News, Mary Ann Azevedo, 21 January 2026, checked 16 October 2026
A study in JAMA (April 2026, five US medical centers, more than 1,800 clinicians with an AI scribe against 6,770 without one): the AI scribe cut documentation time by 16 minutes per 8 hours of appointments and added 0.5 visits a week. Clinicians who used it in at least half of their visits spent 21.3 fewer minutes in the electronic health record per 8 hours of appointments.
JAMA, 'Changes in Clinician Time Expenditure and Visit Quantity With Adoption of Artificial Intelligence-Powered Scribes: A Multisite Study', 1 April 2026, checked 16 October 2026
In the first half of 2026 fintech startups worldwide raised $28.6 billion, almost 23% more than a year earlier, while the number of deals fell to 1,605. Investors make fewer but larger bets and more often choose AI and financial infrastructure.
Crunchbase News, Mary Ann Azevedo, 15 July 2026, checked 16 October 2026
In June 2026 Ramp, a corporate card and expense management service, raised $750 million at a $44 billion valuation. The company reported annual revenue above $1 billion and said its in-house AI tool writes more than two thirds of its code.
GIC, Ramp press release, 4 June 2026, checked 16 October 2026
In June 2026 Taktile, a platform of AI agents for credit, insurance, and anti-fraud decisions, raised $110 million in a Series C led by Goldman Sachs Alternatives. According to the company, the platform helped clients cut false positives in anti-money-laundering checks by 75%.
Taktile, press release, 22 June 2026, checked 16 October 2026
A survey of financial firms by the Bank of Russia, the country's central bank (report of November 2025): 21% use AI on an ongoing basis, 15% are testing pilot projects, 29% plan to start within three years, and 35% neither use it nor plan to. The main barriers are high implementation cost (59%) and a lack of expertise (46%).
ComNews, Pavel Korolev (Bank of Russia survey), 24 November 2025, checked 16 October 2026
In the first nine months of 2026 Venture Guide recorded no venture deals in Russian fintech.
ComNews, Anna Shvetsova (Venture Guide data), 5 October 2026, checked 16 October 2026
The global market for industrial AI was $43.6 billion in 2024 and, by IoT Analytics' forecast, will grow to $153.9 billion by 2030 (23% a year). Yet spending on industrial AI still equals 0.1% of manufacturers' revenue.
IoT Analytics, Knud Lasse Lueth, 9 September 2025, checked 16 October 2026
In June 2026 London-based PhysicsX, which builds AI models for the engineering design of parts from jet engines to semiconductors, raised $300 million at a valuation of about $2.4 billion. Singapore's Temasek led the round, and Nvidia took part.
Bloomberg, Mark Bergen, 8 June 2026, checked 16 October 2026
In September 2026 Japan's CADDi, an AI platform for drawings and manufacturing data, raised $114 million at a $1.2 billion valuation. The company says more than half of Japan's 100 largest manufacturers use it.
Fortune, Jeremy Kahn, 15 September 2026, checked 16 October 2026
Siemens' plant in Nanjing has deployed more than 50 AI applications and digital twins. Compared with 2022, order lead time fell by 78%, time to market improved by 33%, productivity grew by 14% by 2024, and failures at customers dropped by 46%.
Siemens, press release, 15 January 2026, checked 16 October 2026
Rockwell Automation survey (May 2026, 1,560 respondents in 17 countries): AI already supports 34% of manufacturing operations, but only 43% of the data collected is used effectively.
Rockwell Automation, press release on State of Smart Manufacturing 2026, 19 May 2026, checked 16 October 2026
In September 2026 Cognition, the developer of the AI coding agent Devin, raised $2 billion at a $48 billion valuation, almost double its May valuation ($26 billion). Its annual revenue grew from $492 million to almost $900 million in four months.
Reuters, Utkarsh Shetti (republished by The Star), 8 September 2026, checked 16 October 2026
In January 2026 Decagon, a developer of AI agents for customer support, raised $250 million and tripled its valuation to $4.5 billion.
Bloomberg, Shirin Ghaffary, 28 January 2026, checked 16 October 2026
In September 2026 Harvey, an AI platform for lawyers, raised $550 million at a $15.5 billion valuation. It is used by 80% of the 100 largest US law firms.
Harvey, company blog, 9 September 2026, checked 16 October 2026
KPMG survey (July–August 2026, 314 leaders of US companies with revenue of $1 billion or more): 62% of companies are building or deploying AI agents, against 53% a quarter earlier, and the share of multi-agent systems grew from 6% to 25%. Companies most often name higher productivity as the effect (55%).
KPMG, AI Quarterly Pulse Survey Q3 2026, 24 September 2026, checked 16 October 2026
CB Insights survey (February 2026): 80% of companies make AI agent adoption a priority, but 40% cannot measure the return or do not know it. They most often measure the effect in productivity (63%), cost savings (58%), and time saved (58%).
CB Insights Research, 18 February 2026, checked 16 October 2026
Method: how we check the numbers
Every figure on this page comes from an open primary source: a report, a press release or a business media article that cites a study. Every line has a link, a publication date and a check date.
We do not edit an old figure. When a new one comes out, it gets a new line, and the guide’s pages switch to it, so you can see what was true on the check date.
We mark preliminary data as such. For example, the Venture Guide figures for the first 9 months of 2026 are given as of 4 October 2026 and may change.
The check procedure and the reasons sources disagree are described in detail in the guide’s methodology.
We update this page once a month, together with the whole guide. If you find an error or an outdated figure, write to us: we will fix it and set a new check date. The whole section starts from the guide’s main page.