STM · STMicroelectronics N.V.

Current Thesis

The story that mattered three months ago "cheap cyclical auto/industrial semi waiting on a bookings inflection" is no longer the story. Since the Q1 2026 print on 2026-04-23, STM has been repriced as an AI-data-center and space-infrastructure supplier, and the tape reflects it: shares ran from the low-$30s in April to a 25-year high close of $79.71 on 2026-06-03 before easing into the low-$70s. The trigger was the 2026-06-02 decision to roughly double the 2026 data-center revenue ambition to ~$1B (from >$500M) and lift the 2027 target to ~$2B (from >$1B), anchored by a multi-year AWS engagement (2026-02-09), a NVIDIA "Physical AI" collaboration (2026-03-16), high-volume PIC100 silicon-photonics production (2026-03-09), and the decade-old SpaceX/Starlink RF franchise (>5B chips shipped, set to double by 2027). The narrative is genuinely accelerating, but the entry math has changed: this is a momentum-continuation read after a ~2.5x move, not the pre-discovery entry that existed in February-March.

Bull Case

• Data-center revenue ambition roughly doubled to ~$1B for 2026 and ~$2B for 2027 on 2026-06-02 the raise itself, not just the level, is what re-rates a name; it signals visibility improving quarter-on-quarter.
• AWS multi-year deal (2026-02-09) reportedly carries warrants that could give Amazon ~3% of the company a hyperscaler taking equity is the strongest possible demand-durability signal a supplier can get.
• Silicon-photonics PIC100 in high-volume production (2026-03-09) plus power semis for AI server power delivery put STM into two of the highest-growth AI-infra sockets (optical interconnect, power) where competition is thin versus the GPU layer.
• Q1 2026 (2026-04-23): revenue $3.10B, +23.0% YoY (+21.4% ex-NXP MEMS), and a Q2 guide of $3.45B at the midpoint +11.6% sequential, +24.9% YoY confirming the cyclical trough is behind and the recovery is compounding.
• Sell-side catch-up is broad and recent: Mizuho Outperform PT raised to $84 (2026-06-03, from $68), B of A Neutral PT $83 (2026-06-02), UBS to €80 clustered upgrades into strength, the confirmation pattern for an accelerating theme.
• Cluster confirmation across the AI-infra complex: HPE blowout Q2 and Marvell strength (2026-06-02) show the data-center buildout reading through to component and optical suppliers, not just GPUs.

Bear Case

• The cheap re-rating is spent. At low-$70s against PTs of $83-84, the implied upside is ~15-20% the move from skepticism to consensus has already happened, and sell-side now agrees with the tape rather than leading it.
• Margins are still recovering, not strong: Q1 GM was 33.8% (non-GAAP 34.1%) and the Q2 guide of ~34.8% bakes in ~100bps of unused-capacity charges. GAAP Q1 net income was only $37M ($0.04/sh). The multiple is now pricing a 2027 outcome, not a 2026 one.
• AI-data-center revenue is still a single-digit-percent slice of a ~$13-14B revenue base; a ~$1B 2026 target is real but the valuation leans on the $2B+ 2027 trajectory holding any hyperscaler capex digestion (AWS) cracks the thesis.
• Legacy headwinds have not vanished: China SiC wafer ASP pressure and automotive MCU price competition (Infineon AURIX, onsemi Treo) still cap blended margins; the AI story is overwriting these, not eliminating them.
• Stretched and post-spike: an ~11% fade from the 2026-06-03 high after a parabolic catalyst-driven run is how stocks digest a crowded entry; buyers at the low-$70s are paying near 25-year-high prices with the next hard catalyst (Q2 print) weeks away.
• ADR-versus-Paris-ordinary (STMPA.PA) arb spread remains a purely technical downside vector for the US line if EU flow reverses.

Setup & Price Structure

Price structure is a clean uptrend that just printed its first sharp catalyst-spike-and-fade. The 2026-06-02 data-center raise broke the stock to a 25-year high ($79.71 close 2026-06-03), and it has since pulled back into the low-$70s roughly -11% off the spike. The $80 area is now the breakout pivot and overhead resistance; the post-June breakout base sits around $65-68, with the rising 20-week EMA tracking up through the low-$60s. The constructive continuation setup is a higher-low hold of the breakout base and a reclaim of $80 on volume, which would resume the trend; a failure that loses the low-$60s on a weekly closing basis would mark the move as a one-and-done re-rating rather than a multi-leg run. Note the asymmetry has compressed: with PTs at $83-84 only ~15-20% above spot, the risk/reward for a fresh entry at the current level is materially worse than it was for the February-March discovery window. This is strength-as-setup, but late strength size to a continuation trigger, not to the spike.

Catalyst Calendar (next 30 days)

• ~2026-07-23 (est.) Q2 2026 earnings. Just outside the 30-day window, but it is the next hard binary: the print needs data-center revenue visibly tracking toward the ~$1B 2026 target and a Q3 guide extending the +25% YoY trajectory.
• 2026-H2 (date TBA) management has flagged a data-center capacity-expansion deal "expected by year-end" (2026-06-02 commentary); any pre-announcement would be a positive tape catalyst.
• Rolling additional sell-side PT revisions likely to continue clustering after the 2026-06-02/06-03 Mizuho/BofA/UBS moves; watch for any Outperform-to-Buy upgrades that would extend the re-rating, or a downgrade-to-valuation that would cap it.
• No confirmed hard catalyst inside the next 30 days; the tape between now and the late-July print is news-flow and momentum-driven, not event-driven.

What Would Change Our Mind

The thesis breaks on a weekly close below the rising 20-week EMA (low-$60s) that forfeits the post-June breakout base that converts the June run from a new trend leg into a completed re-rating to be faded. Fundamentally, it breaks if the Q2 2026 print (~late July) shows data-center revenue not tracking to the ~$1B 2026 ambition, or if a hyperscaler (AWS) signals AI-capex digestion that undercuts the $2B 2027 path. A China-SiC or auto-MCU margin reset on the call that pulls blended gross margin back below the low-30s would also re-open the legacy-cyclical discount. Conversely, the bull case extends and the limited PT cushion stops mattering only on a clean reclaim of $80 with the data-center capacity deal confirmed.

Correlation Notes

STM now trades as an AI-data-center component/optical name first and a cyclical analog/MCU name second, so its beta has migrated toward the AI-infra complex: Marvell, Broadcom, and the optical-interconnect/photonics cohort (and, by read-through, hyperscaler capex from AWS/Microsoft/Google) drive the marginal move more than the SOX cyclical or auto-semi peers (Infineon, onsemi, NXP) that used to set its tone. The SpaceX/Starlink franchise adds a loosely-correlated space-connectivity sleeve that trades on satellite-buildout headlines rather than the chip cycle. EUR/USD remains a structural margin input STM reports in USD on a largely EUR cost base, so euro weakness is a tailwind. The AWS equity-warrant relationship ties a slice of sentiment to Amazon's AI-infra narrative specifically.