Hashrate orders are booming, why is "AI trading" falling out of favor? JPMorgan reveals the time lag between AI investment and productivity realization

According to APP Zhitong Finance, as global tech stocks—especially in the AI computing power and semiconductor sectors—suffered sharp declines and volatility, Wall Street financial giant JPMorgan Chase released its latest research report stating that AI capital expenditure has already converted into demand for chips, servers, and power equipment. However, the productivity dividend, which could push long-term cost reductions and efficiency increases across the entire economy, has not yet fully materialized. The core message of this report suggests that upstream core chip and key component suppliers involved in the global AI infrastructure boom can continue to receive orders and drive profit growth, while the broader tech sector, including emerging cloud powers and major cloud platforms, simultaneously face valuation pressures from higher interest rates and prolonged return cycles.

The JPMorgan Global Economic Research team, led by Joseph Lupton, published a report on November 10, titled "Productivity Boom: The Contribution of AI Is Not as Big as It Seems." This report extends the team's previous macro outlook: robust economic growth, sticky inflation, long-term higher interest rates, and future AI-driven productivity transformation. The report reveals a time lag between AI productivity gains, the realization of AI dividends, and AI capital spending in the trillions of dollars, implying that widespread productivity benefits may arrive later than the market expects.

After AI equipment is deployed, further steps such as application development, data integration, process restructuring, and employee training are necessary to fully unlock productivity gains. This process may take longer than consensus market expectations. The report cites the "Computer Revolution Paradox" of 1987—"computers are everywhere except in the productivity statistics"—and the subsequent improvements in productivity in the 1990s, illustrating the time lag inherent in technological diffusion.

Fulfilling orders for suppliers, profitability for computing power operators, and broad productivity gains from enterprise AI adoption are three different realization processes. When financing costs rise and the market shortens its patience for returns, strong orders can easily coexist with pressure on equities and bonds. JPMorgan's research provides a macroeconomic explanation for this "AI investment return lag." The report notes that conservative academic research predicts AI will boost total factor productivity (TFP) by less than 0.5% cumulatively over the next decade, while more optimistic studies estimate that AI could raise annual productivity growth by 0.8 to 1.3 percentage points.

After CoreWeave Sounds Credit Alarm, JPMorgan Highlights “Return Lag”

Even as computing power orders continue expanding, the credit market is asking: how much capital do enterprises need to invest, and how long do they wait to see enough cash returns to cover financing costs? CoreWeave recently reported a revenue backlog of $132.5 billion, but its net interest expense in the third quarter totaled $846 million, significantly exceeding its $207 million in operating profit. This vividly demonstrates the tension between growing demand and financing burdens.

Before NeoCloud leader CoreWeave released its latest earnings, the credit market had already sounded alarms about AI investment returns potentially falling far short of expectations, or even that rising debt pressure could force CoreWeave into bankruptcy before realizing strong cash flows. From September 25 to October 26, CoreWeave's (CRWV.US) five-year credit default swap (CDS) spread climbed from 835 to 1148 basis points—an implied cumulative five-year default probability of around 62% under a 40% recovery assumption (this metric also incorporates a credit risk premium and does not directly equate to the company's actual probability of default over the next five years). Meanwhile, Meta's same-period CDS spread rose from 98 to 121.

CoreWeave's CDS spread narrowed to 1041.4 basis points on October 27, showing some easing of pressure; by November 5, its five-year CDS soared above 1289 basis points before retreating. These numbers highlight how the historically high 10-year US Treasury yield is raising the cost of capital for all, and that credit spreads are further stratified by cash flow, leverage, and refinancing needs.

JPMorgan’s Joseph Lupton noted in his November 10 research that recent productivity improvements have been more a result of capital deepening and greater utilization intensity of production factors. Widespread efficiency dividends from AI, through process reengineering, data integration, and organizational adjustments, still require time to be realized.

This largely means that orders for chips, servers, and power equipment can grow first, but sustained investment returns for computing power operators and end-users do not necessarily arrive in sync. In an environment of rising interest rates and widening credit spreads, this waiting period itself increases financing burdens and lowers the present value of future returns. Incremental production and income during the AI construction phase flow first to chipmakers, server manufacturers, and related equipment providers—explaining why shares of leading Asian AI infrastructure manufacturers have been so strong in recent years.

If increasing stock correlations further trigger crowded unwinding of dispersion trades, market adjustments may be amplified. Whether AI assets can regain premiums will depend on how quickly orders convert into cash, whether financing capacity supports project delivery, and whether end-user cost reductions and efficiency improvements are gradually realized.

Chip Rebound Collides with Rate Hikes—Market Rethinks AI Investment Returns

On the US East Coast, at the close on November 5, the Nasdaq 100 plunged 6.36%, the largest daily drop since March 2020, and market cap of its components evaporated by about $2.3 trillion; the Philadelphia Semiconductor Index tumbled 12.25%. After Democrats gained control of both chambers of Congress, the market is worried about increased AI regulation and data center construction constraints, potentially delaying project deliveries, equipment purchases, and revenue realization. Policy concerns triggered valuation adjustments, and systematic strategies such as trend following pared positions, further amplifying the sell-off.

On November 9, pessimism was partly corrected as the Nasdaq 100 and SOX rebounded 3.40% and 6.85% respectively; Micron and SanDisk rose 6.25% and 6.16%. However, Nvidia fell about 4.3% on company-specific news. By Tuesday, November 10, Micron and SanDisk had given up about 1.1% and 3.0% respectively, showing that capital remains torn between robust industry demand and macro valuation pressure. Notably, this recent tech-stock pullback was milder—the VIX dropped from 19.92 to 19.46—indicating that the market is recalibrating Fed rate-path risks, not repeating last Thursday’s panic selling.

On November 10, US CPI data showed overall and core CPI for October rising 0.4% and 0.3% month-on-month, exceeding expectations of 0.3% and 0.2%; the actual core reading met the threshold for "considering a 50 basis point rate hike in December" set by FOMC voting member and Fed Governor Waller. The Dow, S&P 500, and Nasdaq Composite subsequently closed down 0.30%, 0.38%, and 0.74%. Pricing in the market shows the probability of a 50 basis point hike in December has risen to 11.8%, with a 25 basis point increase still as the main expectation. Traders also added bets on three more 25 bp hikes by mid-2027, including the expected December hike. What the market fears is that this round of Fed tightening may be faster and last longer.

Higher Output Per Hour Does Not Mean AI Has Already Delivered a Full-Fledged Efficiency Revolution

JPMorgan first differentiates three concepts that are often conflated: Capital Deepening, Factor Utilization, and Total Factor Productivity (TFP). More equipment and software per worker can boost hourly output; running equipment more intensively and reducing workflow idleness can also lift output; but only after stripping these effects away, does greater output with the same input—true broad-based efficiency improvement—emerge.

Capital deepening itself can also include technological advances embedded in new equipment, so this breakdown should not be interpreted as meaning AI has contributed nothing. For instance, more advanced AI chips improve computational efficiency, with part of that effect captured under capital deepening. Thus, JPMorgan believes recent productivity improvement comes chiefly from capital deepening, which does not mean AI has made no contribution, but rather that AI has not yet—via widespread application and business process reengineering—fully manifested as visible TFP gains.

Key figures show that US labor productivity grew at an annualized rate of 2.6% over the past three years, far higher than the 1.4% during the previous expansion. According to the decomposition cited in the report, capital deepening explains about half of this, factor utilization about 20%, and TFP about a third—TFP grew about 0.9% annually over the past three years. Combined, technology and non-tech capital deepening contributed about 1–1.5 percentage points, with data center-related investments alone possibly contributing as much as 0.5 points. These shares indicate that recent productivity gains have relied heavily on increased capital and more intensive use of existing inputs.

More importantly, the incremental trend matters: the report notes, TFP growth adjusted for utilization has stagnated in the past year. Data updated by the San Francisco Fed on November 9 offers similar evidence—over the four quarters ending in the third quarter of 2026, unadjusted TFP grew 1.27%, but after adjusting for utilization, fell 0.22%. This supports the judgment that "broad-based efficiency dividends have not yet fully appeared," though this does not mean individual enterprise AI adoption offers no benefit.

From the perspective of enterprise deployment, buying GPUs or accessing large models only completes part of the investment. Productivity returns also require usable data, application development, permissions management, workflow redesign, and employee training. These supporting investments come first, with process gains materializing later—forming the "Productivity J-Curve" emphasized by research analysts. As a result, growth in computing power procurement, growing cloud service revenue, and efficiency improvements at end-user enterprises are interrelated but realized on different timelines.

Suppliers Receive Orders First, But the Economy Must Still Overcome Cost and Interest Rate Hurdles

JPMorgan divides technological booms into two phases: investment and return. In the first phase, AI capital expenditure and asset price appreciation drive initial consumption demand, with capital goods suppliers reaping revenue gains first; in the second, broad adoption of technology expands supply, reduces unit costs, and actual income and corporate profits find more durable support—in other words, the first stage sees investment and consumption expansion, with demand outrunning supply improvements; the second is when technology diffusion, productivity growth, and unit cost reduction take hold.

Data in the report shows global capital spending has grown at an annualized 7% since early 2025, with US tech capex up about 30% in the past year. Heavy equipment imports have offset part of the investment’s contribution to US GDP, with manufacturing gains flowing largely to Asia. According to the report, global manufacturing output has grown 5.6% since 2024, with Asia contributing 4.7 percentage points (excluding China and India in the cited charts).

Recent supply chain disclosures align with this timeline. Microsoft's October 28 earnings call revealed quarterly capital spending of $45.2 billion, with about three-quarters used for short-lived assets such as CPUs and GPUs, including higher memory and storage costs. TrendForce noted on November 4 that server DRAM contract prices are still rising, HBM4 volumes are ramping up, and long-term contracts account for a larger portion, even as spot market activity remains subdued. Price hikes and revenue for suppliers mean higher procurement costs for cloud vendors; the well-being upstream can outpace downstream realization of investment returns.

This also explains why improved productivity statistics have not immediately relieved inflationary pressures. Using equipment and personnel more intensively can drive up overtime, maintenance, and wear costs; new capital expands supply, but demand during the construction phase may grow faster. JPMorgan further points out that rising investment and financing needs, alongside a wealth effect encouraging consumer spending, could require a higher neutral real rate to attract savings and support capital formation. Notably, JPMorgan is discussing equilibrium rate changes, meaning that even if inflation falls in the future, the financing environment may not quickly return to previously low rate conditions.

JPMorgan's research also highlights a point easily overshadowed in the AI narrative: consumption cannot be sustained solely by rising share prices and declining savings; expansion must eventually be taken up by wage income. The firm expects hiring to rise cyclically, but weak labor supply could make inflation more stubborn.

Reasoning from this framework, JPMorgan's report shows that future investment across the AI industry chain should focus on whether orders can be converted into cash, whether customers can continuously finance, and whether end-user efficiency improvements can drive the next round of purchases. Profit growth for storage, power, and interconnection suppliers can offset some valuation pressures; platforms and application companies that can turn AI investment into auditable income or cost savings are better positioned to support long-term valuations. Growth driven solely by capital increases will wane as expansion slows, and its contribution to economic growth will diminish accordingly.