Nvidia has committed roughly $40 billion to artificial intelligence investments in 2026 so far, according to reports surfacing this week, before the year is even halfway through. The capital is flowing into customers, suppliers, partners, and rivals all at once. OpenAI is reportedly one of the largest beneficiaries. Other recent disclosures include a $3.2 billion stake in Corning, the optical-fibre maker, and a $2.1 billion investment in the data-centre operator IREN. Nvidia participated in 67 venture deals last year alone, making it one of the most active corporate investors in AI worldwide.
Wall Street's reaction has been to ask whether the shares can run further. Wells Fargo this week raised its price target on Nvidia from $265 to $315, implying a further 44 percent upside on top of an 18 percent year-to-date rally. The analyst, Aaron Rakers, called Nvidia "one of the most attractive secular growth stories in large-cap semis," citing its Blackwell platform and a projected AI data-centre pipeline exceeding $1 trillion by 2027. Fifty-seven of the 61 analysts who cover the stock rate it a buy.
The mood is straightforward. The questions worth asking are not.
Nvidia's investment activity has, until recently, looked like a chipmaker doing strategic deals to grow its market. Now it looks like something else. When the dominant supplier of an essential input takes equity stakes in its largest customers, the line between supplier and shareholder gets blurry. When that supplier also invests in companies that supply its customers, in companies that compete with its customers, and in companies that build infrastructure those customers depend on, the picture starts to resemble something more concentrated than a healthy ecosystem. It starts to look like a holding company that happens to also make the chips.
Some of this is rational from Nvidia's perspective. AI capital expenditure has reached levels that strain even the largest balance sheets. If OpenAI signs a $40 billion compute contract with Nvidia and uses Nvidia's investment to help pay for it, both parties get what they want and the cash effectively cycles. The pattern is familiar from previous bubbles, where vendor financing kept demand running long after underlying economics had become hard to defend. It is not necessarily wrong. It is worth noticing.
The other story this week is what Nvidia is buying outside the model labs. The Corning investment is about optical interconnects: the high-speed cables and switches that make modern AI training possible. The IREN investment is about land, power, and physical data centres. Nvidia is moving downstream into the surrounding stack with the same energy that it once put into building software ecosystems around its chips. The CUDA strategy of the 2010s, where Nvidia made its hardware sticky by owning the developer toolchain, is being reproduced at the level of buildings, fibre, and electricity.
For the AI labs themselves, the relationship is becoming harder to characterise. OpenAI, Anthropic, and others negotiate with Nvidia over chip allocation, lead times, and pricing. They are also, in many cases, recipients of Nvidia capital and tenants in data centres that Nvidia partly owns. The negotiating leverage that comes with that picture is not symmetrical, even before factoring in Nvidia's near-monopoly on the chips that actually train state-of-the-art models. AMD's MI300X is real and improving, and Google, Amazon, and Meta all build custom silicon. None of that has dented Nvidia's grip enough to matter at the frontier.
Antitrust regulators have so far taken little interest. That is partly because Nvidia is American, and the bipartisan consensus in Washington is that American AI dominance is a national priority. It is partly because the investments are individually modest relative to Nvidia's roughly $5 trillion market capitalisation. And it is partly because nobody yet quite knows what to call what is happening. There is no clean precedent for a hardware vendor that bankrolls its industry's largest customers while also owning slices of the infrastructure those customers run on.
The Wells Fargo note is probably right about the next few quarters. Demand is exceeding supply. Margins are durable. The pipeline is real. The longer-term question is whether the AI economy that is emerging is one where many companies compete in a market, or one where a single company sits at the centre of nearly every transaction, alternately wearing the hat of supplier, investor, lender, and landlord. The current trajectory points firmly at the second.