Nvidia adds $277bn in a day

Nvidia added $277 billion to its market value on 22 February, more than the entire worth of most companies in the S&P 500. The shares rose sixteen per cent after another blowout earnings report. No company had gained so much in a single session before, and the move hit portfolios far beyond the chip sector.

The record it broke was only three weeks old. Meta had added close to $200 billion on 2 February after its own results. Nvidia's quarter justified some of the excitement: revenue in the three months to January reached $22.1 billion, more than triple a year earlier, most of it from chips sold into data centres, and the company guided to about $24 billion for the current quarter.

How one chipmaker got so heavy

Nvidia's rise has been fast even by the standards of the past year. Its market value has now climbed close to $2 trillion, and this month it overtook Amazon and Alphabet to become the third-most valuable American company, behind only Microsoft and Apple. A handful of large technology stocks accounted for most of the S&P 500's gains last year, and Nvidia was the biggest winner among them. Any stock that grows that fast grows into the indices with it, because their weights follow market value.

The demand behind it comes from a short list of buyers. Microsoft, Alphabet, Amazon and Meta are racing to build computing capacity for generative AI, and Nvidia's graphics processors, together with the software that programmers have built around them over more than a decade, are the default choice. Supply has been the constraint rather than demand, which has let Nvidia charge prices that produce margins more typical of software than of hardware.

One stock, every portfolio

The jump rippled through quantitative funds around the world. Momentum traders rushed to cover shorts, and systematic long-short funds had to cut their gross exposure. When one company carries this much weight in the indices, its own business risk becomes market risk for everyone else. Concentration has gone so far that diversification now depends mostly on demand for GPU computing.

The effect crossed borders the same day. The S&P 500 closed at a record, and in Tokyo the Nikkei 225 finally surpassed the closing high it set at the end of 1989, helped by Japanese suppliers of chipmaking equipment. European indices also reached records. A single American earnings report moved benchmarks on three continents, which is a measure of how much of global equity returns now rests on one industry's spending plans.

The mechanics of a factor shock

The transmission runs through positioning rather than opinion. A market-neutral fund that is long a basket of cheap stocks and short an index future is, in effect, short Nvidia, because the index contains it. When Nvidia jumps sixteen per cent, the short leg loses money while the long leg barely moves, and the fund's risk models demand a smaller book. Selling the longs and buying back the shorts at once spreads the shock into stocks with no link to chips.

Active managers who are underweight feel it differently. Many diversified funds cap single holdings for risk or regulatory reasons, so they cannot match the index weight of the biggest names and lag their benchmark on days like this one. Momentum funds, which tilt towards recent winners, gain on the day but become even more dependent on a single company. Either way, a set of portfolio exposures that looked varied on paper turns out to share one driver.

Who else rides the wave

The winners extend along the supply chain. Taiwan Semiconductor Manufacturing, which makes Nvidia's chips, and the memory makers that supply the high-bandwidth chips packaged alongside them have gained, as have server builders and makers of chipmaking tools. Electricity utilities and data-centre landlords are increasingly treated as AI plays too. The losers are investors who hedged the wrong way and companies whose own businesses offered no AI story, which have seen capital drain towards the theme.

Precedents and the counterargument

Concentration of this kind has appeared before. The "Nifty Fifty" growth stocks dominated American portfolios in the early 1970s before a painful derating, and in March 2000 Cisco was briefly the world's most valuable company on the strength of demand for internet equipment. Both episodes ended with investors discovering that a crowded bet on one theme had become most of their risk.

The fair objection is that Nvidia is no Cisco of 2000. Its surge rests on revenue that has more than tripled in a year, and on profits that have risen even faster, so the price has partly been chasing earnings rather than hope. That is a reasonable defence of the valuation. It is no defence of the concentration, since a portfolio can hold a fairly valued stock and still be dangerously dependent on it. Orders from a small number of cloud giants drive much of Nvidia's sales, and any pause in their spending would hit the whole market through the same channels that lifted it.

Other risks sit outside Nvidia's control. American export controls, tightened last October, bar the sale of its most advanced chips to China, which was once a large market. Its biggest customers are designing their own AI chips, from Google's tensor processors to Amazon's and Microsoft's in-house designs, precisely to reduce their dependence on a single supplier.

The signals to watch are the capital-spending plans of the largest cloud companies, Nvidia's next quarterly results, and index weightings at each rebalancing. Signs of new competition in AI chips, or of customers designing their own, would test how much of the price is secure.

A $277 billion day has made Nvidia a risk factor in its own right. Until its index weight shrinks, any fund that owns the market is also betting on AI chip demand, whether it means to or not, and the cloud giants' spending plans for this year will decide which way that bet pays.