AI Split the Market in Two

GOOD MORNING, Thursday made the AI divide impossible to miss. IBM and ServiceNow sold off sharply as investors worried AI could compress traditional software growth, while Texas Instruments and other chipmakers rallied on stronger data-center demand. After the bell, Intel added another layer: AI inference is reviving demand for CPUs that many investors had written off. The same technology is creating pressure in one part of tech and scarcity in another. Meanwhile, Brent climbed above $105 and Treasury yields moved higher as Hormuz tensions flared again. AI is still the growth story. Oil is still setting the hurdle rate.

MARKETS | TLDR

Wall Street pulled back from records: The Dow fell 0.37% to 49,309.33, the S&P 500 lost 0.41% to 7,108.30 and the Nasdaq dropped 0.89% as higher oil, rising Treasury yields and software weakness offset strength in semiconductors.
Tech split sharply beneath the index: The S&P 500 Software & Services index fell more than 4%, while the Philadelphia Semiconductor Index rose about 2% to a record, highlighting a widening winner-loser divide inside the AI trade.
Oil returned above $105 Brent: Brent settled 3.10% higher at $105.07 a barrel and WTI rose 3.11% to $95.85 as Iran demonstrated control over the Strait of Hormuz and hopes for a near-term peace deal weakened.

AI BIFURCATION

AI Is Creating Winners and Losers Inside Tech

IBM fell 10.3% and ServiceNow dropped nearly 15% Thursday even though both companies reported quarterly revenue and profit above Wall Street expectations. IBM's software growth slowed, while ServiceNow said Middle East disruptions delayed some deals. The selloff spread to Microsoft, Adobe, CrowdStrike, Intuit and Datadog as investors returned to a broader concern: AI may automate or commoditize parts of the software stack faster than incumbents can monetize it.

The contrast with semiconductors was stark. Texas Instruments surged 10% after forecasting second-quarter revenue and profit above estimates on strong data-center chip demand, lifting analog peers and pushing the Philadelphia Semiconductor Index to a record. The iShares software ETF was down roughly 16% for the year, while the semiconductor ETF was up more than 43%.

That divergence is becoming the central AI market question. The same corporate AI budget can create immediate demand for chips, memory, networking and power while forcing software vendors to defend pricing, seats and workflows. UBS strategist Kiran Ganesh told Reuters the range of outcomes inside tech is likely to widen. "AI exposure" is no longer enough as an investment category; investors increasingly want to know whether a company sells the scarce inputs or competes with the automation those inputs enable.

AI INFERENCE

Intel Found a New AI Tailwind in the CPU

Intel reported stronger-than-expected first-quarter results and gave a sales forecast above estimates as demand for server CPUs used in AI services accelerated sharply. The company said customers had been buying enough central processors that supply tightened and Intel even sold some older finished-goods inventory it had previously expected not to move. The report sent shares sharply higher in after-hours trading.

That matters because the AI hardware story has been dominated by GPUs used for training large models. Inference — serving those models to millions of users — can require a broader mix of processors optimized for cost, latency and power efficiency. If CPUs regain share in that workload, the AI boom becomes materially broader than the Nvidia-centered training cycle and gives Intel, AMD and Arm another path into data-center growth.

The caveat is that part of Intel's upside came from unusually favorable inventory and pricing conditions. CFO David Zinsner later warned that some of the benefit would not repeat in the second quarter. The bigger signal is still useful: AI demand is strong enough to revive products investors had assumed were structurally losing relevance. Now Intel has to prove that inference demand can remain durable after the one-off inventory help fades.

HEADLINES

American Airlines cuts its 2026 profit outlook: The carrier expects its fuel bill to rise by more than $4 billion this year, turning the Iran conflict into a direct margin hit even as travel demand remains resilient.
Honeywell says the war is already hitting revenue: Management estimates Middle East disruptions reduced Q1 revenue by 0.5% and could cut Q2 sales by roughly 1%, showing how logistics and energy costs are moving from macro headlines into industrial guidance.
CenterPoint raises the AI power bet: The utility has more than 12 GW of committed industrial load and plans to serve 8 GW of Greater Houston projects by 2029, with data-center demand pushing its long-term capital plan to about $65.5 billion.
Terafab gives Intel a foundry foothold: Tesla plans to use Intel's 14A manufacturing process for its proposed Terafab AI chip complex, giving Intel a symbolic first major customer for a technology central to its foundry turnaround.
Business activity improves, but prices jump: S&P Global's flash PMI showed U.S. activity accelerating in April while output prices rose to their highest since July 2022, reinforcing the uncomfortable mix of resilient growth and war-driven inflation.

UPCOMING

Microsoft, Alphabet, Amazon and Meta earnings — April 29: The hyperscalers will determine whether the semiconductor strength is being supported by enough cloud growth and AI monetization to justify another year of enormous capex.
Fed decision — April 29: With oil above $100 and business price indicators rising, the Fed's language on inflation and future easing will matter more than an expected hold.
Q1 GDP — April 30: The advance estimate will show how strong the economy was before the full impact of the Iran energy shock moved through household and business activity.
March PCE inflation — April 30: The Fed's preferred inflation measure will test whether energy and supply-chain pressure is beginning to spread beyond headline prices.

DEEP INSIGTHS

AI Deepens the Divide Between Software and Chips

Read this for the clearest current snapshot of AI's uneven economics. Software names are being discounted for potential disruption while chipmakers are being rewarded for scarcity and infrastructure demand — a useful framework for understanding why "AI exposure" increasingly produces very different market outcomes.

CenterPoint's Data-Center Power Buildout

CenterPoint shows where AI capex goes after the servers are ordered. More than 12 GW of committed industrial load and a $65.5 billion infrastructure plan make electricity transmission and generation a measurable part of the AI investment cycle, not just a future bottleneck discussed by chip companies.