Higgsfield, an AI video startup, said it has attracted investment from Nvidia and the growth fund TCV, Bloomberg reported on 23 October 2026, and the announcement is part of Nvidia Chief Executive Officer Jensen Huang’s visit to Kazakhstan, where the company has the majority of its engineering talent. The same day, an Nvidia blog post said Firebird has secured a U.S. export license for 45,000 Nvidia Grace Blackwell GPUs in Kazakhstan, and TechCrunch reported a robotaxi agreement between inDrive, Pony.ai and Nvidia that coincides with Huang’s first visit to the country.
In plain words, a GPU is the chip that trains and runs AI models, and a U.S. export license is the permission Washington requires before the most advanced American AI chips can be shipped to many countries. Sovereign AI is the term for a country building computing capacity and AI companies on its own soil. Neither Higgsfield nor Bloomberg disclosed the size of Nvidia’s investment.
Business Pill · SOVEREIGN AI
A short Business Engineer explainer on sovereign AI: who controls the model and the data. It teaches the general idea only and says nothing about any company in this piece.
The key insight: As we read it, the three same-day announcements sit at different stages: a U.S. export license for 45,000 Grace Blackwell GPUs (Nvidia’s blog), 880 GPUs already deployed by Kazakhtelecom (our arithmetic on the blog’s figures), and an investment of undisclosed size in a startup whose engineers are mostly in Kazakhstan (Bloomberg).
Update (23 Oct): Huang in his own words. At a fireside chat in Astana with Zhaslan Madiyev, Kazakhstan’s deputy prime minister and minister of artificial intelligence and digital development, Huang said of the country: “It’s going to be a hub, a computing hub, an AI factory hub for all of Europe,” CNBC reported. On Higgsfield he said: “The first thing and the last thing is you invest in people, not companies.” CNBC also reported that Higgsfield has more than 300 people working across its engineering and creative teams in Kazakhstan.
Update (23 Oct): banned B300s turn up on a China Telecom network. Lava, an Israeli security firm, says it identified 32 Nvidia Blackwell Ultra B300 chips within the network of state-owned China Telecom, chips barred from export to China by US controls, Tom’s Hardware reported, citing a report by The Information. According to Tom’s Hardware, Lava found them through Nvidia’s own DCGM Exporter monitoring tool and cannot confirm where in the network they sit. An Nvidia spokesperson told the outlet that governments, researchers and media “have all found insignificant diversion of controlled US products, less than one half of one percent in recent court cases — a drop in the bucket compared to the ocean of domestic compute China already has,” adding that “smuggling is a nonfactor.” Tom’s Hardware notes Nvidia downplayed the report but did not deny the findings. It lands the same day as the Kazakhstan licence covered in this piece: the licensed route and the leakage are two sides of the same export regime. Background: the Taiwan B300 indictment in August.
Higgsfield said it has attracted investment from Nvidia and TCV “to help fund global infrastructure and hiring to capture growing demand,” according to Bloomberg’s report.
The round. The new funding is part of an extension to its Series B round in August, which raised $400 million at a $5.4 billion valuation led by DST Global, Bloomberg reported. “Higgsfield didn’t disclose the latest valuation or the size of the extension.”
Users. Higgsfield said its registered users have grown to more than 40 million, according to the report, and that it had added 10 million since its summer fundraising; the report does not say whether that counts users or paying customers.
Where the engineers are. The company is based in San Francisco and has the majority of its engineering talent in Kazakhstan, Bloomberg wrote. It was founded in 2023 by former Snap executive Alex Mashrabov and Kazakhstani AI engineer Yerzat Dulat.
Why TCV invested. TCV, which previously backed Netflix, Spotify and ByteDance, said Higgsfield has shown “a radically faster, professional-grade and AI-native mechanism for creators, brand marketers and agencies alike to turn ideas into images and video”.
Higgsfield serves as a platform to access about 35 AI video models from other providers along with its own AI engine, according to the report. Its U.S. rival Runway raised $315 million at a $5.3 billion valuation in February, which also had Nvidia participating, and China’s Kling AI, owned by Kuaishou, raised $2.8 billion at about a $15 billion valuation this summer, Bloomberg reported.
Nvidia’s 23 October blog gives Firebird a U.S. export license for 45,000 Grace Blackwell GPUs in Kazakhstan; Kazakhtelecom has deployed 880 GPUs (400 H200 + 480 B300, our arithmetic) and plans more than 1,000 GPUs in Data Center Valley and more than 1,000 Vera Rubin GPUs in Ekibastuz next year.Kazakhtelecom’s deployed and planned GPUs and Firebird’s licensed GPUs, as Nvidia’s 23 October blog gives them; 880 and 2,000-plus are our sums of its figures.
What Was Posted on X
.@JensenHuang’s visit comes as @FirebirdCloudAI and Kazakhtelecom expand NVIDIA-powered computing infrastructure, with Firebird planning a Kazakhstan lab to help local startups and researchers build on it.
Nvidia’s own post on 23 October: Jensen Huang’s visit comes as Firebird and Kazakhtelecom expand Nvidia-powered computing infrastructure, with Firebird planning a Kazakhstan lab to help local startups and researchers build on it. It is the company’s first-party account of the visit.
A countryโs data carries its culture, history, and values. It belongs to its people.
@JensenHuang on why that knowledge is a natural resource for Kazakhstanโs AI future.
Nvidia’s account posted that a country’s data “carries its culture, history, and values” and “belongs to its people”, and pointed to Jensen Huang on why that knowledge is “a natural resource for Kazakhstan’s AI future”. Why it matters: it is how Nvidia itself frames sovereign AI for Kazakhstan.
A snipped of my interview with @JensenHuang :
Jensen Huang says countries cannot fully outsource their AI.
He compares it to electricity, the internet and telecommunications: critical infrastructure that a country must be able to keep running.
Rely entirely on outside providers, and you risk losing access to services you depend on.
For Europe, that makes building its own AI capabilities a question of control.
As you know, I sat down with Jensen in Stockholm to discuss sovereign AI, open models and Europe’s next steps.
kimmonismus shared a clip from an interview with Jensen Huang in which, per the post, Huang says countries cannot fully outsource their AI and compares it to electricity, the internet and telecommunications as critical infrastructure.
The Data Point: Nvidia’s Kazakhstan Footprint, by Status
Nvidia’s blog post says “Kazakhstan’s AI ecosystem is moving from infrastructure plans to deployment.” Its figures, with the two same-day announcements, sort into licensed, planned, deployed and invested. By our arithmetic, the GPUs Kazakhtelecom has deployed (880) are about 2% of the 45,000 covered by Firebird’s license.
Who
What the source says
Status
Source
Firebird, with Kazakhtelecom
A U.S. export license for 45,000 Nvidia Grace Blackwell GPUs; first phase under construction at Data Center Valley
Licensed; commercial launch targeted for the third quarter of 2027
Nvidia blog
Firebird, second phase
Total potential AI infrastructure capacity in Kazakhstan of approximately 500 megawatts
Planned
Nvidia blog
Kazakhtelecom
400 Nvidia H200 and 480 Nvidia B300 GPUs in Astana and Almaty, 880 in all by our arithmetic
Deployed
Nvidia blog
Kazakhtelecom
More than 1,000 GPUs in Data Center Valley and more than 1,000 Vera Rubin GPUs in the Ekibastuz region
Planned for next year
Nvidia blog
Higgsfield
Investment from Nvidia and TCV, as an extension of the August Series B ($400 million at a $5.4 billion valuation, led by DST Global)
Announced; size and new valuation not disclosed
Bloomberg
inDrive, Pony.ai and Nvidia
A memorandum of understanding to explore Pony.ai robotaxis on inDrive’s app, with Nvidia’s Hyperion platform and Halos safety system
Tests in Astana in the first half of 2027, with a safety operator
TechCrunch
Sources: Nvidia blog “Kazakhstan Expands Compute for Its AI Future” (23 October 2026); Bloomberg (23 October 2026, as syndicated by The Edge Malaysia); TechCrunch (23 October 2026). The rows have different scopes (licensed, planned, deployed, invested) and are not added together.
The Visit
Kazakhstan’s Ministry of Artificial Intelligence and Digital Development announced the visit on 23 September, according to The Times of Central Asia. Huang was scheduled to hold a fireside chat with Deputy Prime Minister Zhaslan Madiyev at the alem.ai International Center for Artificial Intelligence in Astana, closing the country’s AI Month program, the outlet reported.
The Times of Central Asia also reported that Kazakhstan announced agreements worth up to $10 billion in June connected to the Data Center Valley project in Ekibastuz, which Firebird is developing with Nvidia technology, and that the project is planned to include a computing cluster of around 100,000 advanced GPUs. It said the first Data Center Valley facility is scheduled to begin operating in June 2027; Nvidia’s blog gives a commercial launch target of the third quarter of 2027 for Firebird’s first phase. We have not reconciled the two dates.
The Robotaxi Agreement
inDrive, Pony.ai and Nvidia signed a memorandum of understanding to explore bringing Pony.ai’s robotaxis to inDrive’s platform, TechCrunch reported. Testing in Astana is set for the first half of 2027, inDrive Chief Technology Officer Yuri Misnik said, according to TechCrunch, which reported that the tests will initially involve a safety operator. “Those savings are particularly relevant in emerging markets, where ride-hailing is already relatively affordable and autonomous services need to compete on cost,” Ali Kani, Nvidia’s vice president of automotive, told TechCrunch, referring to Nvidia’s Hyperion platform.
Connecting the Coverage
Our Higgsfield funding piece reported on 19 August 2026 that Higgsfield raised $400 million at a $5.4 billion valuation, four times its January 2026 valuation of about $1.3 billion. What’s new: Nvidia and TCV have joined that round through an extension, at an undisclosed size.
Our Higgsfield run-rate piece reported on 28 September 2026 that chief executive Alex Mashrabov said on the 20VC podcast that the company spends more than $4 million a month on AI models and has crossed $1 billion in annualized revenue, figures we labelled as his own unaudited claims.
Our Vera Rubin piece reported on 21 October 2026 that Nvidia’s release that day said Vera Rubin NVL72 systems “deliver 30x higher throughput per megawatt” for agentic workloads. Kazakhtelecom’s plan for next year includes more than 1,000 Vera Rubin GPUs, according to Nvidia’s blog.
Our SK Hyper piece reported on 23 October 2026 that SK Hyper’s chief executive put the first 5 gigawatts of a Korean data-center plan at $350 billion including equipment, about $70 billion per gigawatt by our arithmetic. Nvidia’s blog sizes Firebird’s potential capacity in Kazakhstan at approximately 500 megawatts.
The Structural Read
As we read it, the Nvidia blog separates GPUs that are licensed, planned and deployed, and the gap between the first and the last is large: Kazakhtelecom’s 880 deployed GPUs are about 2% of Firebird’s 45,000 licensed, by our arithmetic.
As we read it, Higgsfield gives the visit an application-layer name next to the infrastructure figures: a company Bloomberg says has the majority of its engineering talent in Kazakhstan, now with Nvidia among its investors.
As we read it, the robotaxi agreement is the earliest of the three: a memorandum of understanding to explore tests, which TechCrunch says are set for the first half of 2027 with a safety operator.
Nvidia blog, ‘Kazakhstan Expands Compute for Its AI Future’, 23 October 2026
“Kazakhstan’s AI ecosystem is moving from infrastructure plans to deployment.”
Three Implications
FOR CHIP SUPPLY Firebird’s first phase with Kazakhtelecom at Data Center Valley targets a commercial launch in the third quarter of 2027, according to Nvidia’s blog.
FOR AI STARTUPS Higgsfield said the new funding will help pay for global infrastructure and hiring, Bloomberg reported; the amount was not disclosed.
FOR TRANSPORT Pony.ai would provide its Gen-7 robotaxis and Nvidia its Hyperion platform and Halos safety system under the proposed plan, TechCrunch reported.
The framework breaks 2026 compute capex into buyers: “$725B Big Four + $50B Oracle + $13B Apple + $60B Neoclouds + $80B China + $60B Sovereigns + $48B Other”.
As we read it, Kazakhstan’s licensed and planned GPUs belong in the sovereign line of that map. The framework was published in May 2026 and does not discuss Kazakhstan.
What Is Not Established
The size of Nvidia’s and TCV’s investment in Higgsfield and its new valuation were not disclosed. We read Bloomberg’s report as syndicated by The Edge Malaysia and did not read a Higgsfield release. We did not read a transcript of Huang’s remarks in Astana as of 07:30 UTC on 23 October. Firebird’s 45,000 GPUs are covered by an export license; the Nvidia blog does not say how many are installed. We did not contact Nvidia, Higgsfield, Firebird, Kazakhtelecom or inDrive. Nothing here is investment advice.
Business Engineer Framework
The AI Capex Map
A Business Engineer analysis of how the AI build-out is financed, where it cracks and who can absorb it.
On the day of Jensen Huang’s first visit to Kazakhstan, Higgsfield said Nvidia and TCV invested in it, at an undisclosed size. Nvidia’s blog listed a U.S. export license for 45,000 Grace Blackwell GPUs for Firebird, beside 400 H200 and 480 B300 GPUs Kazakhtelecom has deployed (880 by our arithmetic). inDrive, Pony.ai and Nvidia agreed to explore robotaxi tests in Astana, which TechCrunch said are set for the first half of 2027.
A note on sourcing. This piece is based on Bloomberg’s 23 October 2026 report (its free opening and The Edge Malaysia’s syndication), Nvidia’s 23 October blog post, TechCrunch’s 23 October report, The Times of Central Asia’s 24 September report, and our own earlier pieces. The size of the Higgsfield investment was not disclosed. Nothing here is investment advice.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
