Intels AI Comeback|The CPU ratio no one saw coming

The CPU Reversal

Intel just posted its best earnings in years — and the reason has almost nothing to do with Intel.

The company reported first-quarter revenue of $13.58 billion, beating consensus by over a billion dollars. Adjusted earnings came in at $0.29 per share against estimates near breakeven. The stock surged 23% Friday, blowing past its dot-com era peak for the first time in 26 years. But the number that actually moved markets was buried in the data center line: CPU-to-GPU ratios are shifting.

For three years, the AI trade ran on one assumption — GPUs do the work. Nvidia's chips train the models, run the inference, power the cloud. CPUs were legacy. Intel was legacy. That logic held until agentic AI started scaling.

Agentic AI doesn't just answer questions. It plans, delegates, coordinates, and executes sequences of tasks autonomously. That kind of workload moves enormous amounts of data between steps. Intel's CEO Lip-Bu Tan told analysts: customers are deploying server CPUs alongside accelerators "in a ratio that is moving back toward CPU." His CFO added that in certain agentic workloads, the GPU-to-CPU ratio can actually flip. That is a structural shift in the AI hardware market, not a quarterly blip.

Data center revenue at Intel surged 22% year over year to $5.1 billion. AI-related businesses now make up 60% of total revenue, growing 40% annually. The broader chip index — the Philadelphia SE Semiconductor Index — extended its record winning streak to 18 consecutive days of gains, up more than 47% this year. AMD jumped 13% on Intel's coattails. Arm Holdings surged 15%. The read-through: if Intel's CPU thesis is real, the entire data center compute stack needs repricing.

The forward guidance reinforced the case. Intel expects Q2 revenue between $13.8 and $14.8 billion — again ahead of estimates. The CPU business is now projected to post double-digit growth in 2026, a dramatic reversal from expectations of only slight growth six months ago. Supply is actually constrained, which means pricing power is rising alongside volume. Two Wall Street firms upgraded Intel the same morning, with Evercore ISI lifting its price target to $111 from $45.

If agentic AI adoption accelerates as enterprise software companies integrate autonomous agents into production — and early deployment data suggests that is happening — the CPU supercycle could extend well into 2027. But Intel is not yet GAAP-profitable, its GPU ambitions are nascent, and input cost inflation in memory and substrates poses a second-half risk the CFO explicitly flagged. The turnaround is real. The runway is not yet certain.

The $40B AI Bet

The same week Intel's earnings confirmed that AI infrastructure demand is accelerating, Google moved to ensure it controls the supply side of that demand.

Alphabet announced it will invest up to $40 billion in Anthropic, the AI lab behind the Claude models. The deal includes an immediate $10 billion cash injection, valuing Anthropic at $350 billion. This follows Amazon's $5 billion top-up to its own Anthropic stake, with the possibility of an additional $20 billion. Together, the two largest cloud platforms have now collectively committed over $45 billion to a single AI startup — one that had $9 billion in annualized revenue in late 2025 and crossed $30 billion just four months later, a 230% surge.

That number matters for more than Anthropic. It tells you how fast enterprise demand for frontier AI is compounding. Anthropic isn't growing because of consumer chatbot subscriptions. It's growing because companies are embedding its Claude models into software development, legal workflows, and automated coding pipelines at scale. The enterprise pull is structural, not cyclical.

For Google, the deal solves a strategic contradiction. Anthropic is technically a competitor — its models run on Google's own cloud infrastructure. By deepening the investment, Google secures a high-volume anchor customer for its TPU chips and data centers while simultaneously staking a claim in the AI lab most likely to challenge OpenAI's enterprise dominance. The infrastructure synergy is direct: as Anthropic's compute needs grow toward 1 gigawatt of power by year-end, Google's data centers get filled.

The power demand side of this equation triggered its own market event Friday. X-Energy, an advanced nuclear reactor company backed by Amazon, raised $1.02 billion in its IPO — the largest nuclear public offering on record — pricing at $23 per share, $4 above the top of its range. The offering was 15 times oversubscribed. Shares opened at $30.11. The company's xe-100 small modular reactor has an order pipeline of more than 11 gigawatts, with customers including Amazon, Dow, and Centrica. Amazon is separately working with X-Energy to deploy 5 gigawatts across the U.S. by 2039.

The connection is direct. AI data centers require around-the-clock, carbon-free power at a scale that wind and solar cannot reliably deliver. Nuclear is the only clean baseload option at gigawatt scale. The X-Energy IPO wasn't a bet on nuclear energy in isolation — it was a bet on the permanence of AI compute demand.

The risk to this chain is twofold. Anthropic's valuation at $350 billion prices in continuous hypergrowth and sustained enterprise pricing power. If model commoditization accelerates — and DeepSeek unveiled a new competitive model at rock-bottom prices Friday — the margins supporting that valuation compress. On the nuclear side, X-Energy has yet to break ground on a single reactor, and its permit-to-operations timeline runs six to seven years. Patient capital is required.

The Warsh Variable

While chips rallied and nuclear IPOs broke records, the Federal Reserve's leadership succession quietly resolved its biggest obstacle — and the implications for rate policy are anything but resolved.

The Justice Department dropped its criminal investigation into Fed Chair Jerome Powell on Friday, ending a probe that had been blocking a Senate confirmation vote for his designated successor, Kevin Warsh. The investigation — which focused on $2.5 billion in building renovation cost overruns — had been publicly opposed by Senator Thom Tillis, who had pledged to withhold his confirmation vote until the case closed. With Pirro's office shutting the probe, that blockade is lifted. Warsh's confirmation could move quickly enough for him to chair the June FOMC meeting.

A judge handling the case had already ruled the government produced "essentially zero evidence" of wrongdoing. The probe was widely understood as pressure on Powell to cut interest rates — a goal Powell openly resisted. The Fed has held rates steady at approximately 3.6%, citing the Iran war's inflationary effect on energy prices. Trump has publicly demanded rates at 1%, a level no Fed policymaker supports.

The market read this news through the lens of June. If Warsh chairs the June meeting, the question is whether he would cut before the committee is ready — or face dissent from a Fed board that has signaled patience. Analysts noted that Warsh's voting record is historically hawkish under Democratic administrations and dovish under Republican ones. RSM's chief economist put it plainly: Warsh would likely be outvoted by committee members on his first meeting if he pushed for a cut, regardless of his personal inclinations. The actual rate path depends less on Warsh's preferences and more on whether Iran war inflation pressures stabilize before June.

Treasuries gained Friday on the DOJ news, reflecting the market's tentative read that political friction at the Fed is easing. Consumer sentiment data released the same day told a different story: the University of Michigan's index dropped to a record low in April, with inflation expectations rising sharply. Households are pricing in sustained energy-driven price pressure. That is exactly the data the Fed uses to justify holding rates.

The weight of evidence points toward a delayed rate cut cycle — the Fed's own communications have held steady, and Warsh's institutional constraints will limit any unilateral move. If Iran ceasefire talks — Pakistani Foreign Minister Araghchi traveled to Islamabad Friday for preliminary discussions — produce a real de-escalation and oil prices retrace, that calculus changes fast. The verification benchmark is oil: if West Texas crude moves back below $85, June rate cut odds should reprice within days. If crude holds above $90 and CPI prints hot again in May, a June cut becomes nearly impossible regardless of who chairs the meeting. The scenario that would prove this analysis wrong is a Warsh confirmation moving slower than expected — buying the current committee more time to build consensus before any political pressure can arrive at the table.