Peter Thiel has cleared out his entire Nvidia stake during Q3, according to filings reported by Reuters and Bloomberg. Thiel Macro sold 537,742 shares in the chipmaker. Reuters said the holding would have been worth about $100 million based on Nvidia鈥檚 closing price on 30 September. Bloomberg described the sale as a retreat from the company that has become the most valuable name in AI hardware.
The sale came during a period filled with talk about an AI bubble. Reuters said traders have grown uneasy after SoftBank sold its own Nvidia stake last week. Bloomberg also pointed out that Michael Burry has taken bearish positions against Nvidia and Palantir, adding more tension in a market already filled with questions about valuations.
Thiel Foundation did not comment when Reuters asked for a response. A representative for Thiel also declined to comment when Bloomberg made the same request.
听
How Does Thiel鈥檚 Portfolio Look After The Exit?
听
According to the filings, Thiel Macro now holds Apple, Microsoft and a smaller Tesla position as its main long term bets. This leaves the fund with exposure to companies that sit at the centre of global tech but without direct involvement in the highest profile AI chipmaker.
Bloomberg said Thiel has backed AI in other ways. His venture firm, Founders Fund, invested in OpenAI in March 2025 at a valuation of $300 million. He has also supported US based Substrate, which hopes to compete in AI chip production, as well as Mercor and Cognition AI.
SoftBank made its departure from Nvidia earlier in October when it sold shares worth $5.83 billion to help finance other AI ventures. Bloomberg noted that both Thiel and SoftBank founder Masayoshi Son walked away from Nvidia near the time it crossed the $5 trillion mark.
听
Are Investors Divided On AI Stocks?
听
Nvidia鈥檚 share price has gained about 1% since the end of September. Bloomberg said the stock went down as much as 3% on Monday. An analysis of 909 hedge funds found an even split during the quarter. A total of 161 funds increased their Nvidia positions while 160 cut back.
Bloomberg said Burry posted a short message on X saying 鈥渟ometimes, we see bubbles鈥�. He then questioned accounting choices made by hyperscalers such as Microsoft and Alphabet, making the argument that longer depreciation schedules have helped keep expenses down and profits up.
听
Are We Closer To The AI Bubble Burst Than We Think?
听
Experts share their thoughts on whether this move is an indication of the AI bubble burst鈥�
听
Our Experts:
听
- Rajive Jain, Investor & Coach
- Lukman Otunuga, Senior Market Analyst, FXTM
- Ryan Dolley, Vice President of Product Strategy, GoodData
- Kenn van Hauen, Chief AI Officer, AND Digital
- Stuart Harvey, CEO, Datactics
- Simon Noble, CEO, Cezanne
- Oscar Asly, Group CEO, M4Markets
- Petr Baudis, Co-Founder and CTO, Rossum
听
Rajive Jain, Investor & Coach
听
听
鈥淚 believe some investors may exit positions in AI due to rebalancing moves on their part and to hedge against uncertainty looking ahead. Shares of Nvidia have been on a rocket ship in the last few years and to extrapolate from that rise to another linear rise of the same magnitude is not a rational expectation. And since Peter Thiel retains some stake in Tesla post-sale, it seems there remains a conviction the Tesla story is not entirely over in terms of growth in the share prices. Furthermore, he continues to hold a stake in Microsoft, which remains one of the largest AI players in the AI ecosystem.
鈥淲hile AI has been a hot topic, we see daily investment and joint venture deals between the big players. Just today, we saw another partnership between Microsoft, Nvidia and Anthropic, developers of the Claude AI system. Such partnerships are built to enable the infrastructure that AI computing needs 鈥� namely hardware, software, backend platforms, data center access and development of LLM models as the market catches up to more of the technologies.
鈥淭he AI revolution is resulting in vastly higher price multiples. It now requires assessing returns for allocations in AI to those from utilities and other sectors, the lack of which may result in a corrective course for market multiples.
鈥淚 believe the general hype around AI will bring about definite benefits and deployments in corporate environments. How much the AI stakeholders are able to generate returns on investment and in how much time is being closely watched. There will be an overall tendency to reverse course until the steep curve of adoption moderates to a more stabilizing one as usage catches up to investments.鈥�
听
Lukman Otunuga, Senior Market Analyst, FXTM
听
听
鈥淧eter Thiel鈥檚 Nvidia exit is likely to fuel investor jitters about an AI bubble.
But whether the AI hype will end in a bubble that will eventually burst remains anybody鈥檚 guess.
鈥淕iven sky-high valuations, circular deals, and significant disinvestment, concerns are mounting over the AI bet. However, unlike the dot-com era, the AI sector is tangible with continued robust demand for AI chips supporting the business outlook.
鈥淣evertheless, the bar remains high for Nvidia with investors expecting another round of solid earnings for Q3. Anything less than exceptional may trigger a selloff with options implying a $320 billion swing for the stock in either direction post-earnings.鈥�
听
Ryan Dolley, Vice President of Product Strategy, GoodData
听
听
鈥淚 use AI every single day. It saves me hours and makes me think better. That鈥檚 not going away. Financial bubbles burst when people stop believing in value. But I know the value because I鈥檓 getting it right now. Productivity revolutions don鈥檛 collapse like speculative assets.
鈥淎I has a bubble, but the technology is real. When dot-com crashed, the internet didn鈥檛 disappear, it became infrastructure. Same thing happens here. The froth settles, the tourists leave, and the transformation keeps going.
鈥淲e鈥檙e in AI鈥檚 dot-com moment. Lots of excitement, lots of noise, lots of inevitable failure. What matters in 2026 is speed of learning. The winners won鈥檛 be the ones with the best pitch, they鈥檒l be the ones who figure out what actually works while everyone else is still talking.
鈥淭he dot-com crash took a decade to recover financially, but the internet reshaped everything during that time. It didn鈥檛 wipe out jobs, it transformed them. AI follows the same pattern. Once the hype burns off, the real builders get back to work.鈥�
听
Kenn van Hauen, Chief AI Officer, AND Digital
听
听
鈥淧ichai鈥檚 frank acknowledgment of 鈥榠rrationality鈥� in the current AI boom reflects what many across the industry are seeing 鈥� a period of extraordinary progress mixed with equally extraordinary expectations. Phases like this naturally bring volatility alongside genuine breakthroughs.鈥�
鈥淚f momentum slows or a correction comes, it won鈥檛 signal failure of the technology but a recalibration of how quickly it can be commercialised and scaled. The long-term trajectory remains strong, but the near future will separate companies creating genuine value from those dependent on continued exponential valuation growth.鈥�
听
Stuart Harvey, CEO, Datactics
听
听
鈥淭he businesses that will remain the most resilient throughout the broader AI cycle are the ones treating data readiness and data quality as a core pillar rather than an afterthought. The most powerful models still collapse without clean, well-governed data and transparent processes.
鈥淎s AI adoption accelerates, so does the risk that poor data management will result in unreliable outputs, regulatory issues, or costly rework. A market correction would only amplify the importance of effective governance, because when the hype subsides, the advantage shifts to the teams that invested in trustworthiness, quality, auditability, and responsible development from the start.鈥�
听
Simon Noble, CEO, Cezanne
听
听
鈥淎I is experiencing a period of rapid growth, and with that comes excitement. Some of the hype will inevitably settle as expectations meet reality, but that doesn鈥檛 mean the bubble will burst. The focus will instead shift from experimentation to maturity.
鈥淭ake the recent Workday AI legal case which made the headlines. It highlights why understanding how AI models are trained is so important, especially if they鈥檙e going to stand the test of time. Systems are only as good as the data that shapes them. If the data isn鈥檛 diverse or representative, bias can creep in without anyone intending it.
鈥淭o build AI that lasts and genuinely supports fair, inclusive outcomes, models need to be trained on wide-ranging, balanced datasets; not just the data a company happens to have. Being transparent about data sources and training approaches helps build trust, accountability, and resilience across the industry.
鈥淪o, while the current AI hype may settle, that doesn鈥檛 mean it will burst. Responsible, well-trained AI will continue to grow as the industry matures 鈥� eventually becoming a standard part of how we work.鈥�
听
Oscar Asly, Group CEO, M4Markets
听
听
鈥淭he news that Peter Thiel has exited his stake in Nvidia is striking, but it does not in itself signal an imminent collapse of the AI thesis. What it does indicate is a recalibration of expectations from one of tech鈥檚 most influential contrarians.
鈥淣vidia remains the clearest barometer of AI infrastructure demand. When a seasoned investor steps aside, it suggests he sees valuations running ahead of fundamentals in the near term.
鈥淭hat said, this looks more like a strategic repositioning than a rejection of the broader AI story. The underlying drivers such as enterprise deployment, chip demand, and data centre expansion are still firmly in place. What is more likely is a normalisation phase: less euphoria, more disciplined valuation work, and a clearer divide between firms building real, scalable AI capability and those driven mainly by narrative.鈥�
听
Petr Baudis, Co-Founder and CTO, Rossum
听
听
鈥淎fter leading AI scientists poured cold water on the hype of achieving Artificial General Intelligence soon, most circles of AI insiders are talking about the 鈥淎I bubble鈥� popping soon. But every AI boom ends the same way: with a few systems quietly doing the dull things better than we ever could. That鈥檚 what savvy investors and budget-holders are trying to find right now.
鈥淭he AI bubble is real, but AI foam is a better name for the effect we鈥檙e going to observe on the markets soon. Many companies that sell AI are hoping the tech will catch up with their value proposition, but expectations are (finally) being corrected as apparent progress slows. This correction will make AI bubbles pop 鈥� be it overhyped startups, technology vendor laggards frantically trying to pivot, or even some of the frontier AI lab hyperscalers.
鈥淏ut there is just one thing more harmful than investing in overhyped AI products that fall short 鈥� and that is ignoring specialised AI that already, right now, solves real problems with measurable results.
鈥淗ow do we do that? The AI bubble is overblown with hype about AI agents. AI agents do have huge potential productivity upsides compared to mere human AI assistants. But we hear these agents supposedly do everything within a huge range: from math research to coding, document organisation, running your marketing or finance team, and building software products autonomously. But here鈥檚 the key: what鈥檚 their real-world track record? Not just testimonials, but hard numbers. Are they moving the needle? And it鈥檚 not just about cost savings. If quality drops after saving money, that鈥檚 a problem. How well tested are they?
鈥淭here is a clear pattern in the answers to these questions 鈥� laser-focused specialised agents work better than generalists. Think 鈥� do you want a 鈥済eneralist agent that manages your supply chain鈥�, or 鈥渟pecialist agent that processes incoming invoices and saves the accepted ones to your ERP system鈥�? Specialised agents are to achieve a clear set of tasks, and have guardrails that help them achieve these tasks.
鈥淭hat gives them extra reliability, and even their failure mode, thanks to those guardrails, means they can easily hand off to a human rather than underperform or fail silently. Thanks to these handoffs, you can clearly see the agent鈥檚 efficiency and impact on your bottom line. And you can still solve every problem, because in the worst case, the AI鈥檚 human colleague is still there to take over.
鈥淭his story repeats field over field 鈥� from customer support to financial analysis, from software engineering to scientific research. Being AI-savvy doesn鈥檛 mean just ignoring the hype, but also recognising real business value. The alternative is wasting money and time, and being left in the dust by competition.鈥�
