Nvidia's $60 billion profit draws Japan 1980s comparisons as AI capex bets reach $1.5 trillion
Record chipmaker earnings now face historical skepticism, with four companies projected to spend $1.5 trillion on data centers through 2027.
Nvidia reported net profit roughly doubling from a year earlier to about $60 billion, a number that stood at just $6.2 billion three years ago, according to a Foreign Policy analysis published Friday (2026-08-28). Revenues have doubled accordingly, hitting $96.22 billion.3
The scale puts the chipmaker's earnings power in a category of its own, but the framing around the numbers has shifted. The same analysis draws a direct line to 1980s Japan, when soaring equity valuations and dominant corporate earnings preceded a prolonged downturn. Nvidia is already the world's largest company by stock market value at more than $5 trillion.3
The company said it expects next fiscal year's revenues to rise a further 70 percent. That guidance rests on an extraordinary assumption about capital spending: this year and next, just four companies are projected to spend $1.5 trillion building data centers that will be full of Nvidia's chips — Amazon, Google, Meta and Microsoft.3
Traders weighing the numbers against energy markets face a familiar problem. The data center buildout is a known demand driver for power, but the pace of construction and the actual electricity load remain harder to pin down than the chip orders themselves. Platts JKM LNG front-month closed Friday (2026-08-28) at $23.17/MMBtu, unchanged on the session, reflecting tight Asian LNG balances that a sustained pullback in data center construction could soften.3
The earnings report landed in a week when investors had plenty of other numbers to digest. TotalEnergies reported adjusted net income of $6 billion for the second quarter on Thursday (2026-07-23), up 68 percent from $3.578 billion a year earlier and 12 percent from the first quarter, meeting analyst expectations. Equinor followed on Wednesday (2026-07-22) with $4.84 billion in net income, up 267 percent year on year, and announced a third buyback tranche of up to $1.125 billion.2,1
None of those figures approach Nvidia's scale, which is precisely why the historical comparison carries weight. In the 1980s, Japanese corporate profits and equity valuations marched upward together until they didn't; the Nikkei peaked in late 1989 and spent the next three decades below that level. The Foreign Policy piece argues the current AI rally has similar structural features: a dominant supplier, concentrated buyers, and a consensus that the spending cycle has years to run.3
The counterargument is equally specific. Nvidia's profit base is real, not accounting fiction, and its customers are generating cash flow from the infrastructure they are building. The four hyperscalers projected to spend $1.5 trillion are not speculative startups; they are the same companies that have funded the cloud and internet economy for two decades. Bulls would note that the 1980s comparison ignores the revenue that AI deployments are already producing.3
For energy markets, the distinction matters less than the trajectory. A slowdown in data center construction would hit power demand forecasts in the US, Europe and Asia simultaneously, and the electricity sector has already priced in years of AI-driven load growth. Platts JKM LNG front-month at $23.17/MMBtu reflects tight Asian balances, but a capex pullback from the four hyperscalers would soften that picture quickly.3
The concentration of the AI trade cuts both ways. Four companies account for the overwhelming share of the $1.5 trillion data center pipeline, and their capital allocation decisions are made by a handful of CFOs. One guided reduction in cloud capex guidance would ripple through chip orders, power procurement and LNG contracting within a quarter.3
Nvidia's own guidance of 70 percent revenue growth next fiscal year is the figure to watch. It implies the hyperscalers maintain or accelerate their buildout pace, and any sign of hesitation in their quarterly calls would move the chipmaker's stock and spread quickly across power procurement and LNG markets. Equinor's buyback program (up to $3 billion for 2026, with the third tranche now approved) suggests energy producers see their own cash flows as durable, but that confidence may prove easier to maintain than the spending commitments underpinning AI power demand.1
The 1980s Japan analogy offers a warning, not a forecast. Corporate earnings can double while the market tops out; the two are not mutually exclusive. The next catalyst is the hyperscaler earnings season, when Amazon, Google, Meta and Microsoft will have to justify their share of that $1.5 trillion against the same skeptical lens now being applied to Nvidia.3