STM · STMicroelectronics N.V. · Stock research

Current Thesis

The June AI-data-center leg is now confirmed broken by a second lower low: the reference close of $50.56 on 2026-08-21 sits -36.7% below the 52-week high of $79.83 and below the $54.29 close of 2026-08-14 that the prior note anchored to. The $58 weekly level flagged on 2026-07-19 gave way in July and has not been reclaimed. The shares are down 24.3% over three months.

The narrative is dead, and the dating is now three points rather than two — the 2026-06-22 high of $79.82; the 2026-07-23 Q2 print that beat on revenue, EPS and the data-center ambition and was sold 13% premarket to $57.22; and 2026-08-21, when TrendForce reported a third 2026 price increase effective 2026-08-23 alongside power-device lead times of 52 weeks, and the Euronext line closed at 43.14 EUR, down 0.34%. Operating datapoints keep landing and the multiple keeps compressing through them.

What an investor would be buying at this level is a different proposition from the one that ran in June: an analog/power pricing cycle feeding a gross-margin step — 33.8% in Q1, 34.8% in Q2, 37.0% guided for Q3 — with no company-specific event of consequence until the Q3 print expected 2026-10-22. That thread has not yet been bid.

Bullish and bearish views on STMicroelectronics N.V.

The model's bull view on STMicroelectronics N.V. (STM), in brief: Third 2026 price increase takes effect 2026-08-23 across multiple product lines, following a first hike announced 2026-03-24 (effective 2026-04-26) and a second effective 2026-06-28 (TrendForce, 2026-08-21). The bear view: The forward multiple is not distressed: 25.04x on the 2026 EPS estimate of $1.30 as compiled on 2026-08-21, after a 36.7% drawdown. Both cases follow in full.

Bull Case

• Third 2026 price increase takes effect 2026-08-23 across multiple product lines, following a first hike announced 2026-03-24 (effective 2026-04-26) and a second effective 2026-06-28 (TrendForce, 2026-08-21). Trade press reports automotive MCU prices already up 15–20%.
• Power-device lead times reported generally beyond 30 weeks with some at 52 weeks as of 2026-08-21 — a supply condition that is what makes a third hike in five months enforceable.
• Q3 2026 guidance (2026-07-23): $3.70B revenue at the midpoint, +6.2% sequential and +16.2% YoY, with gross margin guided to 37.0% after absorbing 70bps of unused-capacity charges.
• Q2 2026 (2026-07-23): net revenues $3.49B, +26.0% YoY and +12.7% sequential, 110bps above the guide midpoint; adjusted EPS $0.31 against $0.25 consensus; GAAP net income $222M, diluted EPS $0.24.
• The data-center ambition was raised twice in seven weeks — 2026-06-02 to ~$1B for 2026 and ~$2B for 2027, then again on 2026-07-23 to "above $1B" and "well above $2B."
• Consensus compiled as of 2026-08-21 shows an average target of $75.20 across 21 analysts with a mean recommendation of Outperform, against a 2027 EPS estimate of $2.77 versus $1.30 for 2026.
• The repurchase programme continued through the drawdown: 65,209 ordinary shares bought 2026-08-03 to 2026-08-07 at a weighted average of EUR 45.7643, EUR 2,984,246.06 in total.

Bear Case

• The forward multiple is not distressed: 25.04x on the 2026 EPS estimate of $1.30 as compiled on 2026-08-21, after a 36.7% drawdown. The de-rating has room to continue without the shares becoming statistically cheap on the current year.
• Q3 revenue guidance of $3.70B at the midpoint came in fractionally under the ~$3.72B consensus at the time. The upgrade in the guide was on the margin line; the top line was not.
• Sell-side targets are converging downward while the average lags: Goldman Sachs at $60.50 (cut from $67.50) and Barclays at $60 (cut from $65), against a compiled average of $75.20. That gap is a mechanical source of further negative headlines.
• The company's own July execution is underwater — 96,195 shares repurchased 2026-07-06 to 07-10 at a weighted average EUR 60.27, against a Euronext close of 43.14 EUR on 2026-08-21. The August tranche at EUR 45.7643 is close to the tape and no longer a support argument.
• Price increases pushed into 52-week lead times are the classic set-up for double-ordering; the 2021–22 analog cycle ended with a distribution inventory correction. The 2026-08-21 tape read the hike as a demand risk, not a margin gift.
• Structure has failed at every level the June story rested on: rejections at $79.71 (2026-06-03) and $79.82 (2026-06-22), loss of the $58 shelf in July, $54.29 on 2026-08-14, $50.56 on 2026-08-21.
• The $1.5B convertible issued into the June strength remains an overhang on the share count.

Setup & Price Structure

The reference close on 2026-08-21 is $50.56. RSI(14) reads 45.7 — a fresh low for the move with a mid-range momentum reading, which describes an orderly grind rather than a washout. The one countertrend session of size in August was 2026-08-07, when the shares closed up 5.30%; it did not lead to a higher high.

Crowding and positioning observables, stated as observables:

• Distance from the 52-week high of $79.83 is -36.7%; the three-month price change is -24.3%.
• The compiled average target of $75.20 (21 analysts, as of ~2026-08-20) sits roughly 49% above the 2026-08-21 close — the sell-side base has not re-marked to the tape, and two houses cut inside the last two weeks.
• No earnings date inside 30 days. The next print is expected 2026-10-22, roughly two months out, which removes the event-driven bid entirely.
• Insider or issuance-into-strength evidence: none disclosed in the window. The recurring corporate transaction is the weekly repurchase, whose weighted average has fallen from EUR 60.27 (week of 2026-07-06) to EUR 45.7643 (week of 2026-08-03).
• Retail-sentiment coverage in the last week clustered on the pricing headline rather than on the data-center story that drove June — a shift in what the coverage is about, with no accompanying price response.

Reclaiming $58 on a weekly close would be the first evidence the July break was a shakeout rather than a regime change. Nothing in the 2026-08-21 tape argues that yet.

Catalyst Calendar (next 30 days)

• 2026-09-22 — Ex-dividend date for the $0.09 quarterly installment, payment 2026-09-23. The only confirmed corporate date inside the window.
• ~2026-10-22 (est., outside the 30-day window) — Q3 2026 results. Tests the $3.70B midpoint, the 37.0% gross-margin guide and whether "above $1B" of 2026 data-center revenue is backed by a quantified H2 run-rate.

Elapsed catalysts

• 2026-08-23 — Third 2026 price increase takes effect across multiple product lines. Whether it sticks shows up in distributor commentary and, eventually, the Q4 gross-margin guide. (passed 3d ago)
• ~2026-08-24 (est.), and weekly thereafter — Share-repurchase disclosure on Form 6-K. The only recurring print of where the company itself transacts; the 2026-08-03 to 08-07 tranche averaged EUR 45.7643. (passed 2d ago)

What Would Change Our Mind

The residual case rests entirely on one line: the 220bps sequential gross-margin step to a 37.0% Q3 midpoint. A Q3 print around 2026-10-22 with gross margin under 37.0%, or a Q4 guide carrying reshaping charges beyond the 70bps already flagged, removes the only part of the story still compounding and leaves a de-rating auto/industrial franchise with a small AI line.

The second thing that would settle it is repetition. Good news has now been sold twice — the 2026-07-23 beat-and-raise and the 2026-08-21 pricing/lead-time headline. A third dated positive datapoint met with a negative price reaction would establish that the market is repricing the multiple independently of delivery, at which point the fundamental sequence stops being investable on this horizon.

On price, a weekly close below $48 would break the $50 handle the shares have traded around since the Q2 print and confirm the markdown is still extending. In the other direction, a weekly close back above $58 would reclaim the pre-breakout shelf lost in July and would be the first structural evidence against the failed-breakout read.

A book-to-bill disclosed below 1.0 at the Q3 call, or distribution inventory described as rebuilding after having been called normalized in Q1, would say the pricing actions pulled demand forward.

Correlation Notes

• Dual-listed: the Euronext line closed at 43.14 EUR on 2026-08-21 while the US reference close was $50.56. Reporting currency is USD against a substantially EUR-denominated cost base, so FX lands directly in the gross-margin line rather than below it.
• The 2026 pricing cycle is industry-wide — NXP, Texas Instruments and Infineon have also raised prices this year — so the pricing thread is a sector read, not a company-specific one. Peer confirmation or contradiction arrives at onsemi's Q3 print (~2026-10-26 est.) and Infineon's fiscal Q4 in early November, both of which also set the read on SiC ASPs.
• The data-center line is tied to hyperscaler capex through the AWS multi-year engagement disclosed 2026-02-09; a capex digestion signal from that customer transmits directly to the "well above $2B" 2027 framing.
• Roughly $700M of exposure sits in the Apple time-of-flight socket, which links a slice of 2027 estimates to iPhone 18 supply-chain reporting rather than to the semiconductor cycle.