Chipmakers AI-driven earnings surge

For most of the generative-AI cycle, 'AI capex' has effectively been shorthand for 'Nvidia.' April 23, 2026 is the day that story broke open. Texas Instruments, a 94-year-old analog house best known for op-amps, power-management ICs, and calculators, posted 19% revenue growth and guided Q2 revenue to $5.0-5.4B versus a $4.86B consensus — and the stock jumped 19% for its best single day since 2000. The reason isn't a new GPU architecture. It's that every AI server rack the hyperscalers build needs dense arrays of TI's voltage regulators, signal-conversion chips, and interface parts. TI's CEO Haviv Ilan put a number on it: the data center segment grew 'around 90 percent from the same period last year,' and is now a >$1B annual revenue business layered on top of an industrial base that is itself recovering from a multi-year downcycle.

Intel's print reinforces the same thesis from the other side of the chip. The Data Center and AI segment hit $5.1B, up 22% YoY and ~$700M above consensus, and CFO David Zinsner specifically attributed the beat to 'the growing and essential role of the CPU in the AI era.' That framing would have been laughed off in 2023. In 2026 it's the beat. Put the two earnings together and the market read is unambiguous: AI capex is spilling out of the GPU silo and into analog, CPUs, advanced packaging, and interconnect — the full bill of materials for an AI data center, not just its accelerator. That's why the Philadelphia Semiconductor Index crossed 10,000 for the first time on the same day these two companies reported. The rally is broadening, not narrowing.

Intel shares spent a quarter century trapped under the ~$75.81 all-time closing high they set during the 2000 tech bubble. That ceiling broke this week. Q1 2026 marks Intel's sixth consecutive revenue beat, and non-GAAP EPS came in at $0.29 against a $0.01 consensus — a miss by a factor of 29x on the way up. After-hours trading sent the stock up 15-20%, and YTD performance is now roughly +105%. For a company whose narrative for most of the last decade was 'structurally disadvantaged,' this is a regime change.

Lip-Bu Tan, who joined as CEO in 2025 after his Cadence turnaround, has repeatedly framed Intel's opportunity as the shift from training to inference and agentic workloads. His formulation — 'The next wave of AI will bring intelligence closer to the end user, moving from foundational models to inference to agentic' — matters because it maps directly onto Intel's product stack. Agentic workflows orchestrate tool calls, spreadsheets, browsers, and pipelines; that orchestration runs on general-purpose CPUs, not H100s. If Tan is right about where the compute shifts next, Intel's x86 installed base becomes an asset rather than a legacy. The Q2 guide — $13.8-14.8B in revenue and $0.20 non-GAAP EPS — is well ahead of where sell-side models were set, and multiple brokers have already repriced (Northland to $92, Stifel to $65). The turnaround thesis is no longer a contrarian call.

Read the quiet part out loud: TI's best single day since 2000. SOX's best month since February 2000. Intel breaking its 2000 closing high. Three separate 'since 2000' datapoints in one session is not a coincidence — and Wall Street has noticed. The harianbasis recap flagged that SOX is now trading more than 16% above its 50-day moving average and at a 52-week high, a configuration that historically precedes mean reversion. Year-to-date gains on individual chip names have decoupled from any reasonable growth rate: AMD +242%, Micron +41%, Broadcom +38%, Nvidia +22%, all in a single quarter.

The bull counter is that the fundamentals are arriving, not just the multiple. Bank of America upgraded TXN to Buy with a $320 price target and raised 2026-28 EPS forecasts by 21%, 31%, and 33% respectively — numbers that only make sense if this is a sustained capex super-cycle, not a sentiment spike. BofA separately projects global semiconductor sales to grow 30% YoY in 2026 and cross the $1T annual milestone. Commentary across the week has reframed this phase as an 'industrialization' of AI — meaning deployment at factory scale rather than the speculative generative-AI enthusiasm of 2023-24. The tension for investors is real: the earnings confirm demand, but the calendar comparisons to 2000 are the kind of echo that has cost people money before. Dan Ives of Wedbush states the bull case flatly — 'no cracks in AI demand on the chips/hardware or software front' — but 'no cracks yet' is a sentence that ages in interesting ways.

Intel's move is louder, but TI's is the more analytically interesting result, because TI doesn't make AI accelerators. It sells boring components into boring end markets, and yet its Q1 is the one that crystallizes the broadening thesis. Three numbers do the work. First, analog revenue was $3.92B, up 22% YoY — that's the core franchise accelerating, not just a hot adjacent segment carrying the print. Second, the data center line grew ~90% YoY to a >$1B annual run-rate, demonstrating that even non-accelerator silicon is being pulled into hyperscaler buildouts at a rate that dwarfs any historical compare. Third, the Q2 guide ($5.0-5.4B revenue, $1.77-$2.05 EPS) is more than $500M ahead of the Street at the midpoint, with the industrial segment — long the overhang on this name — inflecting at the same time.

Stifel's Tore Svanberg framed the industrial piece precisely: 'industrial is particularly strong with lean inventories and improving sell-through.' That's the key structural point. TI is not getting a one-quarter AI bump on top of a weak base. It's getting an AI-driven data center acceleration layered on top of an industrial recovery where customer inventories have already been drained. The combination is why BofA's 33% EPS hike for 2028 is survivable — the thesis is that TI is now a dual-engine compounder rather than a cyclical commodity analog shop. For investors trying to size the AI capex wave, TI is a cleaner read than the hyperscalers themselves, because it's priced on delivered orders rather than capex promises. If the wave is real and durable, TI's P&L is where it shows up first without accelerator distortion. Business-news video coverage over the last 24 hours leaned heavily into this exact framing, treating the print less as a company story and more as an industry-wide signal.

The phrase 'AI industrialization phase' appeared in analyst commentary before these prints; the prints are what make it concrete. Industrialization here has a specific operational meaning: AI stops being a concentrated bet on frontier-model training capex at a handful of labs and hyperscalers, and becomes a distributed bet on deployment infrastructure across thousands of racks, edge locations, and enterprise installations. That shift is what creates the next four quarters of non-obvious watch items.

First, inference and agentic deployment creates steady-state capacity growth rather than lumpy training orders — which is why TI can credibly guide Q2 up after a monster Q1 without flagging pull-forward, and why Intel's Q2 guide of $13.8-14.8B implies sequential acceleration rather than mean reversion. Lumpy-to-smooth is the P&L signature analysts should look for in Micron, Broadcom, and Lam Research prints in the next two weeks. Second, capital intensity spreads from accelerators into facility spend: power delivery, cooling, interconnect, and rack-level integration. That is specifically why TI's analog franchise is printing +22% rather than simply its data center segment — the bill of materials has widened. Third, the forward indicator to watch is not revenue, which is already rising, but order book duration. BofA's 33% EPS hike for 2028 makes sense only if orders are being booked with multi-year visibility; if TI or Intel flags shortening order tails in the Q2 call, the 'industrialization' framing weakens fast. The marketminute framing of silicon as 'the world's most critical commodity' is less hype than logistics. Over the next four quarters, the earnings signal is whether order visibility holds — not whether demand is loud today.