Kioxia's Record $11.78B Profit Forecast Highlights the AI Economy's Biggest Divergence

On July 31, 2026, Japanese NAND flash memory maker Kioxia issued a blockbuster forecast for the July–September quarter: operating profit of ¥1.89 trillion ($11.78 billion) . It was the latest milestone in a staggering run that began earlier in the year and has made Kioxia one of the clearest winners of the AI infrastructure buildout.

Yet this same boom is squeezing the biggest AI builders in the world — Amazon, Alphabet, Microsoft, and Meta — who are spending faster than they can generate cash. The result is a structural divergence that defines the AI economy in 2026: component suppliers capture the profit, while infrastructure builders absorb the cost and the valuation risk.

Kioxia's Numbers: Pricing Power at an Extreme

Kioxia's fiscal Q1 2026 (April–June) already shattered records:

• Revenue of ¥1,767.1 billion — up 76% quarter-on-quarter and 415% year-on-year .
• Operating profit of ¥1.27 trillion — roughly 28 times the ¥44.9 billion reported a year earlier .
• Operating margin guided at approximately 74% , a level of profitability rarely seen in the semiconductor industry .

The engine behind these numbers is not rising shipment volumes but pricing power. Average selling prices (ASPs) surged 70% in Q1, even as bit shipment volumes actually fell . This confirms a market in which AI data center demand has created an "AI premium" pricing regime — customers are paying far more per gigabyte of NAND, even as they buy fewer total chips.

Kioxia has also confirmed that NAND supply will remain "extremely strong" and fully booked through calendar 2026 . That means customers cannot get additional allocation at any price. The company's outlook for Q2 projects operating profit of ¥1.89 trillion ($11.78B), a figure that exceeds the full-year operating profit of many established semiconductor companies .

The Hyperscaler Squeeze: Spending $725 Billion

While Kioxia enjoys sold-out capacity and 74% margins, the four largest hyperscale cloud builders are locked in an accelerating spending race.

Their combined capital expenditures for 2026 are on track to reach approximately $725 billion — up 77% from roughly $410 billion in 2025 :

• Amazon: ~$200 billion (up ~100% from 2025)
• Microsoft: ~$190 billion (up ~100%)
• Alphabet (Google): $175–185 billion (up ~110%)
• Meta: $115–135 billion (up ~80%)

The numbers are so large that they are beginning to strain even the most cash-rich companies in history. A Reuters analysis published July 22 notes that while these four companies will generate about $340 billion more in annual operating cash flow in 2027 than in 2025, their capex is expected to rise by roughly $534 billion over the same period — equivalent to about $1.57 of additional investment for every $1 of additional cash flow .

Free Cash Flow: A Rapidly Tightening Vice

The most closely watched metric is free cash flow (FCF), and the trajectory is alarming investors.

Aggregate free cash flow across the four largest hyperscalers is projected to fall to near zero by mid-2026 and may turn negative in Q3 . Operating cash flow is growing at roughly 23% annually, while AI capex is growing at roughly 70% annually — a pace that cannot be sustained indefinitely without external financing or spending cuts .

The pressure is already visible in quarterly filings:

• Alphabet's stock dropped 7% after the company raised its 2026 capex forecast by $15 billion to a range of $190–205 billion. Amazon, Meta, and Microsoft all fell in sympathy as investors scrutinized "dwindling cash piles with uncertain returns" .
• Alphabet's Q2 2026 free cash flow turned negative for the first time in the company's history, with outflows of $5.8 billion .
• Amazon's trailing free cash flow has collapsed 95%, from approximately $38 billion to $1.2 billion .
• Morgan Stanley estimates that hyperscaler FCF will hit its lowest level since 2014 .

Evercore ISI analysts described the situation as a "red flag" for stock valuations, noting that the major AI hyperscalers are spending so much on capex that some may go cash-flow negative .

The Core Asymmetry: Who Wins in a Constraints-Driven Market?

Kioxia's results and the hyperscalers' cash burn are two sides of the same coin.

Memory suppliers like Kioxia sell into the AI buildout but face almost no supply-side competition for their product. NAND is effectively sold out, so pricing power is extreme, margins are spectacular, and profits compound. When a 70% ASP increase flows through to a 74% operating margin, nearly all the upside stays with the supplier.

The hyperscalers buy the memory, GPUs, networking, and power to build AI infrastructure — but they compete with each other, face no pricing relief, and are locked in an arms race that consumes all operating cash flow. Investors are punishing them for it .

This is the defining financial asymmetry of the current AI cycle: the constraints are on the supply side, and the companies that own those constraints capture the profit. The companies that need to buy through those constraints absorb the cost and the valuation risk.

Outlook: Can the Divergence Persist?

Analysts expect combined hyperscaler capex to top $1 trillion in 2027 . Whether that spending generates enough revenue to restore free cash flow growth remains an open question. Schroders notes that while revenues and earnings are increasing, the capital base is increasing much faster, which is having a negative effect on return on invested capital (ROIC) .

Some analysts project recovery by 2028 if AI capacity begins converting into stronger cloud and advertising returns . But for the remainder of 2026, the numbers point in one direction: memory suppliers like Kioxia will continue to enjoy extraordinary margins, while the hyperscalers face intensifying scrutiny on every dollar they spend.

The divergence is not a temporary blip. It is a structural feature of an AI buildout where the most critical components are supply-constrained, and the buyers have no choice but to pay whatever the market demands.