Goldman Sachs warns AI capex will peak in early 2027; big short seller Burry doubles Nasdaq 100 put options
Goldman Sachs' 2027 global market outlook report warns that the accelerating wave of AI capital expenditure will most likely peak in early 2027, with the market shifting from a one-way rally to two-way volatility; its estimates show that cloud providers' capital expenditure's boost to S&P 500 EPS will reach a peak of 12 percentage points in 2026, fall back to 9 points in 2027, and turn into a 5-point drag in 2028. Michael Burry doubled his Nasdaq 100 put options and re-established short positions in semiconductor ETFs, calling this a "bull trap" comparable to the top of the 2000 internet bubble. The Federal Reserve's financial stability report also named artificial intelligence as a potential threat to asset valuations and financial stability.
Body · AI translation
Goldman Sachs recently released its 2027 global market outlook report, warning that the accelerating wave of AI capital expenditure will most likely peak in early 2027, and that the market will shift from a one-way uptrend to two-way volatility, with recommendations to increase diversified allocation and hedging;
The Federal Reserve also explicitly named artificial intelligence in its latest financial stability report, warning that it poses a potential threat to asset valuations and financial stability.
And the well-known short seller Michael Burry chose to make his stance clear directly through positions, doubling his Nasdaq 100 put options while re-establishing short positions in semiconductor ETFs, stating bluntly that this is a "bull market trap," especially benchmarked against the top of the 2000 internet bubble; Jensen Huang's departure further deepened his concerns about the entire AI industry chain.
01
These factors are, in fact, also where recent market concerns lie.
The Goldman Sachs report says that the increment of AI investment as a share of GDP will peak in early 2027, but it does not say that capital expenditure will turn downward—only that the pace of increased investment will slow.
And stock market pricing inherently places more weight on marginal changes; the past bull market in the upstream computing power sector relied on continuously rising capital expenditure growth, and once this driving force stops, the core support for high valuations will loosen.
According to Goldman Sachs estimates, there is a clear corresponding relationship in the curve of cloud vendors' capital expenditure pulling S&P 500 EPS: contributing 2 percentage points in 2024, 5 points in 2025, surging to a peak of 12 points in 2026, falling back to 9 points in 2027, and directly becoming a 5-point drag on earnings in 2028.
And 2026 is basically the final window for capital expenditure to drive earnings.
The step-down in earnings growth also reflects this change in sync; related earnings expectations for the S&P have dropped from 36% to 11%, with European, Japanese, and Asian markets weakening in tandem.
Objectively speaking, 11% growth is not bad by historical standards. What is uncomfortable is the decline from 36% to 11%.
What was previously enjoyed was a Davis double play of earnings and valuations rising together; going forward, most likely only slow earnings support and sustained valuation pressure will remain, and it will be hard to replicate the past two years of easy wins.
The second unavoidable real constraint is the high interest rate market environment.
Cloud giants like Microsoft and Google have deep pockets and abundant cash flow, so the impact is limited; those suffering most are the mid-tier GPU cloud service providers, such as CoreWeave and Nebius, named by Burry. They are major customers of Nvidia, expanding rapidly through large-scale leasing and purchasing of GPUs, and are themselves highly dependent on external financing.
Once interest rates rise, their borrowing costs directly increase, expansion is forced to contract, and since their valuations were originally assigned based on extremely high growth, the expectation gap can easily widen.
The circular transaction risk mentioned by the Federal Reserve is also worth noting.
Tech companies and startups invest in each other, purchase from each other, and endorse each other's revenue; when funding is loose, this can beautify the books and push up valuations, but after financing tightens, it becomes a string of bound risk exposures. If one company has a cash flow problem, the risk will transmit along the cooperation chain. This is not a question of valuation level, but a hidden danger at the credit level.
And the biggest recent negative trigger is the Jensen Huang incident, which shook the pricing anchor of the entire chain.
It is still difficult to determine how much long-term impact this matter will have on Nvidia's own products and orders.
Over the past few years, the order expectations and valuation logic of many companies in chips, foundries, memory, and downstream cloud vendors have been built on Nvidia's strong position. Once this anchor wavers, everyone will reassess the assets in their hands.
The big short seller Burry has clearly seized this window, but his approach has also changed.
Previously he preferred to bet directly on individual stocks such as Nvidia and Palantir; now his focus has shifted to index and semiconductor ETFs. While doubling Nasdaq 100 put options, he is also buying back VanEck semiconductor ETF short positions, and TSMC and Nvidia are precisely the heaviest holdings in this ETF.
He benchmarks the Nasdaq's early October high against the 2000 bubble peak and calls out a "bull market trap," focusing on whether Nvidia's influence will gradually weaken.
Of course, his strategy is not necessarily right; after all, the difference in this analogy is very large.
During the 2000 bubble, many companies had no cash flow and were mired in debt; now the balance sheets of leading U.S. tech giants are quite healthy, and the valuation premium of the five major tech stocks relative to other stocks has narrowed to a very low level, far from the madness of that year.
Moreover, this options bet has a very high threshold: it only becomes profitable if the index falls more than 20% before expiration, so time, magnitude, and direction must all be right.
There is another layer of pressure that is easy to overlook, coming from a regulatory shift.
Put back into the capital expenditure curve, this variable will further lower the slope of expansion. Originally, computing power was expected to continue explosive growth; now, with more approval, energy consumption, and compliance constraints, the peak comes earlier and the realization pace is slower.
With four layers of pressure stacked together, it is only natural that the valuation bubble inflated by optimistic narratives is squeezed.
02
There is also a very interesting signal that Goldman Sachs has mentioned many times.
The forward PE of U.S. cloud giants is only 19.7 times, below the lower bound of their own 20-year historical range, and even lower than consumer staples, which are widely recognized as a defensive sector.
For many people, high-beta growth stocks have seen valuations fall below defensive stocks, which shows that a large part of pessimistic expectations has already been priced in, and not everyone is blindly optimistic.
Tighter regulation is not entirely bad in the long run. Once the disorderly arms race in computing power stops, everyone will be forced to do the math on whether the computing power invested can bring stable revenue and profit.
After the bubble is squeezed out, the market's opportunity structure will also switch to a different logic.
Two years ago, the simplest strategy was to go all in on the upstream shovel sellers—chips, optical modules, storage; whoever was closest to computing power rose the most. That logic was built on continuously accelerating capital expenditure, and it will be hard to replicate going forward.
The market's evaluation criteria will shift from "who is expanding capacity" to "who can earn real money."
The first category worth tracking is the cloud giants whose valuations have corrected sufficiently. Slower capex growth actually means less spending and improved free cash flow for them. With solid balance sheets and diversified revenue structures, they are gradually transforming from high-volatility growth stocks into defensive names with growth attributes, and their cost-effectiveness will gradually become more prominent in a choppy market.
Next, there will be repeated scrutiny over whether AI can help customers save money and increase revenue, and whether there is stable payment and renewal. Only companies that capture real productivity gains and have a clear ROI can potentially take over as the new main theme.
This is also what Goldman Sachs calls the "relay team," and whether the handoff goes smoothly will determine whether this round of adjustment is a soft landing or a continued selloff.
The third category is cyclical sectors that have already priced in a large amount of pessimistic expectations, such as Korean memory. Share prices have already implied expectations of a sharp 70% downward revision to 2027 earnings, yet order schedules are sold out through 2029, and PB has been compressed to historical lows. The odds look high, but the risk is by no means small, and volatility is extremely severe.
Goldman Sachs has also not blindly turned bullish on Asia; its core concern is excessive concentration, with TSMC, Samsung, and SK Hynix accounting for too heavy a share, so if the logic reverses, the drawdown could be very damaging.
Conversely, there are also many names that should be actively avoided. Small and mid-sized GPU cloud providers that are highly tied to Nvidia orders, lack stable cash flow, and rely on financing to stay afloat, as well as pure cash-burning startups with no launched products, are the most fragile groups in this round of stress testing.
03
A more prudent approach is to reduce concentrated bets on a single track and diversify positions appropriately.
The market has now shifted from a one-way bull market to two-way volatility, and the value of hedging and multi-asset allocation will be higher than in previous years.
This round of AI narrative is facing a test—not of whether AI can change the world, but of whether these investments can be converted into stable returns. The process of the bubble deflating is bound to be grueling, and volatility will persist for a long time, but after the market clears out, assets that can withstand the test and have solid cash flow will usher in the next leg of a market that goes further.
During this period, just stay calm, withstand the pressure, and wait slowly for the dawn.
This article is from the WeChat public account "Gelong," author: Duxingxia, published by 36Kr with authorization.