As their stock prices have reached all-time highs, tech giants have doubled annual capital spending to over $200 billion since 2022, driven primarily by AI infrastructure build-out whose returns are not yet visible in earnings. Analysts expect that figure to approach $500 billion annually by 2026–2027¹ and it is still unclear just how much ‘winning’ in AI will eventually cost.
This stands in sharp contrast to the last sustained period of value increase for these companies, around a decade ago, which was fuelled by unprecedented growth and profitability as consumer engagement shifted to mobile and key platforms became near-universal. Then, markets rewarded realised performance. Today, they are rewarding investment well ahead of it — a historical anomaly.
For perspective, the Magnificent Seven alone have added roughly $6 trillion in market value since late 2022, largely driven by expectations tied to AI. While revenue and profit growth for most platforms (conspicuously excluding Nvidia) are in the single or low-double-digit range, the inevitable price of ubiquity is that it becomes hard to grow much faster than the overall economy. Apple as one example is growing at 6% currently, which is the ‘curse’ of already generating $400 billion in revenue. In modern financial history, there has never been such a divergence between actual achievement and the degree of reward.
The net result is that valuations have increased much more today, during a period of very little increase in profitability, than they did a decade ago when there was far more evidence. Without waiting for tangible evidence that this capex is delivering meaningful incremental growth, markets are effectively assuming guaranteed future returns, with virtually no adjustment for time or execution risk. This begs the natural question: is this a simple case of another market bubble?
We believe the answer at its core is no, with a few significant caveats. What is different today from 10-20 years ago is the much-greater understanding and appreciation of network effects. While it has always been the case that winners in any emerging tech segment have accumulated the majority of available profits (Apple, Microsoft and Google are three testaments to this), what is different today is the degree and speed of this accumulation. Precisely because of the last two decades’ experience, investors today understand that only a very small number of players can ‘win’ in AI and more importantly, that the stream of future profits available to any winner are dramatic and persistent. This is a fundamental underpinning behind valuations rewarding sustained capex.
Also, the downstream incremental profit margins available after capex build-out could be near 100%. An imperfect analogy is building a nuclear power plant which provides most of the power to an entire region and therefore operates as a quasi-monopoly. It requires substantial capital upfront, but the marginal cost of output once built is near zero.
Finally, the potential value of ubiquity in the current AI wave is far more extensive than it was with mobile. Two decades in, e-commerce via mobile is ‘only’ about 60% of total e-commerce and less than 10% of total commerce. The mobile platform has become central, but not ubiquitous. AI on the other hand is already on track to become embedded in literally every aspect of consumer life, and the path to enterprise ubiquity already appears clear. This before we have the expected radical improvements in functionality. In fact, the comparison with mobile is only valid in terms of looking at periods of unprecedented valuation ramps; one is a platform shift, the other (AI) is a complete transformation.
AI capex is at its core rational. In a transformational shift of this scale, standing still is not an option for companies at the infrastructure layer, and the scale of that investment simply mirrors the ubiquity of the downstream opportunity. The alternative quite simply is irrelevance. This in fact isn’t so different from the shift to mobile 10+ years ago. Then, Facebook bet the entire company on it. If Facebook had not gone mobile, what would it be worth today?
The caveats are all around the degree of valuation increase, not its direction. At a minimum, valuations would normally adjust for time since many of these investments will not deliver meaningful increases in performance for years. Capex typically reduces the present value of these future returns, because capital must work for longer before delivering profit. This is why capital-efficient businesses have historically been awarded higher multiples: SaaS and internet platforms compared to hardware or infrastructure-heavy companies.
There is also uncertainty, for example in the pace of enterprise AI deployment. It is one thing to ‘vibe-code’ a simple single-thread process, quite another to manage integrated processes across time zones in real time without failure. If enterprise deployment of AI takes longer than expected, new data-centre capacity may not be fully utilised in the near term, and rising energy costs and deployment constraints are real. The future risk that any ‘factory’ (i.e. datacentre) will run part-empty waiting for demand reduces value today. The only question is how much.
This is an entirely rational arms race priced ‘irrationally’ simply because every participant is priced as if they will win. This in turn matters, all down the investing food chain. The top seven technology companies represent around 30% of the S&P 500 and virtually drive the NASDAQ, and so their valuation logic spreads into private markets, venture portfolios, and strategic M&A. This is now a market-wide dynamic, not a 7-company technology narrative.
For growth-stage technology companies, the implication is clear. This environment is creating real valuation air-cover and strategic appetite but simply will not last in its current form. Someone will stumble, something will be late, and something will not go according to plan. This is the curse of being priced to perfection, which many private companies discovered in the wake of raising at too-high valuations in 2021 for example.
Companies positioned as enabling or accelerating this infrastructure cycle will benefit disproportionately. Those that cannot articulate their role in the emerging stack risk losing momentum when markets re-anchor to realised performance.
For founders and boards, this is the moment to be proactive rather than reactive. Sharpen the narrative, validate the commercial model around AI (not just having AI functionality), and time strategic exit processes while the valuation ‘umbrella’ remains favourable. There are many strategic buyers today who are actively shopping to acquire AI capability in already-built companies, across areas as diverse as legal-tech and drug discovery and diagnosis.
Sources
¹ Reuters — Tech spending plans will test stock market’s AI trade, 29 October 2025. Forecasts Big Tech capex approaching ~$500 billion annually by 2027. https://www.reuters.com/business/tech-spending-plans-will-test-stock-markets-ai-trade-2025-10-29/
Victor Basta is Managing Partner at Artis Partners. He has advised growth-stage technology companies on more than 130 transactions over a 30-year career, with a focus on sell-side financings and strategic M&A. He has helped found and build three investment banks, led global advisory practices, taught exit preparation at INSEAD, and is a regular commentator on technology markets and M&A.
Contact Information
For enquiries, please get in touch via our contact page.
Press enquiries: Alicia Burke, Marketing Director — [email protected]
This communication is issued by Artis Partners LLP, an appointed representative of Thornbridge Investment Management LLP, which is authorised and regulated by the Financial Conduct Authority.
The contents of this communication are provided for general informational purposes only and do not constitute financial promotion. It does not constitute investment advice, or an offer or invitation to buy or sell any financial instrument or engage in any investment activity.
No reliance should be placed on the information contained in this communication for any investment decision or activity. All the information in this communication is believed to be reliable but may be inaccurate or incomplete. All rights reserved.
Artis Partners LLP (FRN:1022191) is an appointed representative of Thornbridge Investment Management LLP (FRN: 713859), which is authorised and regulated by the Financial Conduct Authority.
