Everspin’s beat is not a beat: product demand is improving, but the EPS print depends on non-operating support

The print is easy to misread because the headline EPS beat language from management conflicts with the street-comparison basis: EPS was actual $0.06 vs estimate $0.12, a surprise of -47.8%, while revenue was actual $14.1 million vs estimate $14.4 million, a surprise of -2.7%. What was priced in was not a dramatic acceleration, but the sell-side did expect a cleaner EPS outcome than $0.06 against $0.12, and it expected revenue closer to $14.4 million than $14.1 million. What actually surprised on the positive side was not the headline versus consensus, but the mix underneath it: MRAM product sales in the third quarter were $12.7 million, up 22% over the third quarter of last year, while licensing, royalty, patent and other revenue decreased to $1.4 million from $1.7 million in Q3 '24. That matters because the business investors want to own is product MRAM, not episodic project and licensing revenue; Q3 showed that product revenue is carrying more of the load even as EPS missed the street estimate.

That distinction is the thesis, because Everspin’s revenue trajectory has shifted from the 2024 trough into a measured recovery, but operating leverage is still not proven. Revenue fell from $14.4 million in Q1 FY2024 to $10.6 million in Q2 FY2024, then rebuilt to $12.1 million in Q3 FY2024, $13.2 million in Q4 FY2024, $13.1 million in Q1 FY2025, $13.2 million in Q2 FY2025, and $14.1 million in Q3 FY2025. The sequential move in Q3 FY2025 was +6.5%, and the year-over-year move was +16.3%, which is a credible recovery from the -26.6% revenue YoY printed in Q3 FY2024. But gross margin stayed at 51.3% in Q3 FY2025, the same as Q2 FY2025 at 51.3% and below Q1 FY2026 at 52.7% in the quarterly history, so the recovery has not yet translated into a sustained return to the 2023 margin profile of 56.8%, 58.4%, 60.2%, and 58.1%. The right interpretation is therefore not “growth is back” in a generic sense; it is that product revenue has stabilized the top line, while margins remain structurally lower than the 2023 base that supported diluted EPS of $0.04, $0.18, $0.11, and $0.09.

The margin point explains why the EPS miss versus the street matters even though management framed the quarter as inside its own range. Sanjeev Aggarwal’s wording is worth using because it commits to the company’s guidance framework rather than the consensus framework: “We are pleased to report third quarter results with revenue of $14.1 million and non-GAAP EPS of $0.06 per diluted share with revenue in line with our guidance range and EPS towards the high end of our expectations.” That sentence tells the PM where the disconnect sits. Everspin did what it said it would do against its guide, but the market had $0.12 in the model, and the actual street-comparison EPS was $0.06. The accounting bridge is visible in the call excerpt: GAAP operating expenses were $8.8 million, up slightly sequentially and increased from $8.1 million in the third quarter of 2024, while other income of $1.2 million was related to the strategic award won in August of last year. If operating expense is $8.8 million and other income is $1.2 million, the quarter’s $0.06 non-GAAP EPS cannot be capitalized like recurring product operating profit unless Q4 shows the product ramp absorbing that expense base.

The product ramp has three identifiable legs, and the quality of the print depends on which leg does the incremental work in Q4. IBM’s FlashCore Module 4 is not an acceleration story in the near term. Aggarwal said revenue from the sale of PERSYST 1 gigabit STT-MRAM into IBM's FlashCore Module 4 or FCM4 for data center applications “remained consistent with the prior quarter” and that Everspin continues to anticipate revenue from this project “to remain at this level for the remainder of the year.” That language is deliberately constraining: IBM supports the base, but management is not asking investors to model a Q4 step-up from FCM4. Lucid Motors is a different read-through, because Everspin continues to ship and recognize revenue for its PERSYST MRAM solution for the Gravity SUV and expects volumes to increase as the automaker ramps production. The magnitude disclosed for the whole product bucket is $12.7 million, up 22% over Q3 '24, but Everspin did not split IBM from Lucid Motors or LEO satellite customers, so the defensible conclusion is that IBM is steady, Lucid Motors is ramp-dependent, and satellite production must carry the next layer if Q4 is to move meaningfully above the Q3 level.

That satellite leg is the real variant perception because management put a near-term production marker on it, and the street revenue miss may have obscured that the timing is now inside the next quarter rather than an open-ended design-win story. Aggarwal said Everspin “continued to ship engineering samples of the PERSYST EM064LX HR and EM128LX HR to several LEO satellite customers and remain on track to ramp to full production in the fourth quarter 2025.” The important part is not that LEO satellites are an attractive market, which would be too vague to underwrite; it is that Q3 already included engineering samples, the parts are named as PERSYST EM064LX HR and EM128LX HR, and full production is tied to the fourth quarter 2025. If Q4 revenue lands near the top of $14 million to $15 million, the satellite transition likely contributed incremental product revenue. If Q4 merely holds near Q3’s $14.1 million, the “full production” language will have been absorbed by offsets elsewhere or by small initial volumes.

The second-order read-through is narrow but actionable because Everspin names customers and programs even though the supply-chain table lists no customers of MRAM and no suppliers to MRAM. IBM’s FCM4 should be modeled as stable through the remainder of the year, not as an upside catalyst, because the company says the revenue remained consistent with the prior quarter and will remain at this level for the remainder of the year. Lucid Motors gets a supplier-content signal from Everspin’s continued shipments into the Gravity SUV, but the revenue magnitude is only disclosed within total MRAM product sales of $12.7 million, so the read-through is that Lucid Motors’ production ramp can change Everspin mix without giving investors enough data to infer Lucid Motors unit cadence. For the unnamed DoD contractor, the implication is more concrete below the operating line: Everspin recognized $1.2 million in other income in the third quarter and $8.5 million to date from the $14.6 million contract to develop a sustainment plan for MRAM manufacturing facilities to provide continuous onshore MRAM capabilities to aerospace and defense customers. That award supports cash and EPS optics, but it is not the same quality as product gross profit from IBM, Lucid Motors, or LEO satellite customers.

The cash flow and balance sheet soften the downside, but they do not erase the need for operating proof. Everspin ended the quarter with cash and cash equivalents of $45.3 million, up $0.3 million from $45 million at the end of the prior quarter, while cash flow generated from operations decreased to $0.9 million for the third quarter from $5 million in the second quarter. The reason given was higher collections on receivables on a change in distributors, which makes the Q2 cash flow comparison less useful as a run-rate. The cash balance gives Everspin room to fund the elevated expense base, but Q3’s operating expense of $8.8 million against revenue of $14.1 million and gross margin of 51.3% leaves limited room for EPS expansion without either a higher revenue level or richer mix. The tariff point is also cleaner than many semiconductor prints: William Cooper said the company “did not experience any tariff-related impact on our results in the third quarter and do not expect any tariff-related impact in the coming quarter.” That removes one exogenous excuse from Q4, which raises the burden on product execution.

The peer comparison reinforces that Everspin is not a cycle beta trade in the way mainstream memory names are, which is why the market may be applying the wrong mental model. In the Memory peer table, MU shows revenue of $41,456.0 million with gross margin of 84.6% and revenue YoY of +345.7%, while WDC shows $3,337.0 million, gross margin of 50.2%, and revenue YoY of +45.5%, and STX shows $3,112.0 million, gross margin of 46.5%, and revenue YoY of +44.1%. Everspin’s Q3 FY2025 revenue of $14.1 million and gross margin of 51.3% put it much closer to an application-specific specialty supplier than a commodity memory upcycle participant. The consequence is that multiple expansion should follow named-program conversion, not the broader memory group’s gross-margin recovery. Investors buying Everspin because memory peers are printing large revenue YoY gains risk overpaying for a business whose next leg depends on IBM stability, Lucid Motors ramp volume, and LEO satellite production timing rather than generalized DRAM or NAND pricing.

The call delivery supports the same cautious but constructive read because management’s prepared remarks sounded better than the Q&A, and the tone data shows investors did not get uniformly stronger conviction. In the tone history, Q3 FY2025 sentiment was 0.35 versus Q2 FY2025 at 0.38, guidance_tone was 0.27 versus Q2 FY2025 at 0.43, prepared_sentiment was 0.39 versus Q2 FY2025 at 0.51, and qa_sentiment was 0.04 versus Q2 FY2025 at 0.19. At the same time, ai_optimism was 0.93 in Q3 FY2025 versus 0.59 in Q2 FY2025, while uncertainty was 58.4 versus 59.4 and qa_evasiveness was 51.2 versus 28.9. Those numbers conflict in a useful way: the model detects more optimism, but lower guidance_tone and higher qa_evasiveness suggest that management enthusiasm was not matched by cleaner investor-facing specificity in Q&A. For a small-cap specialty semiconductor name, that matters because the next quarter hinges on the exact timing and magnitude of production ramps that management has not broken out by customer.

That tone split also explains why the Q4 guide should be treated as the near-term referendum on the thesis rather than a routine continuation of Q3. Cooper guided Q4 total revenue in the range of $14 million to $15 million and GAAP net income per fully diluted share between $0.02 and $0.07, with non-GAAP net income per fully diluted share between $0.08 and $0.13. Relative to Q3 actual revenue of $14.1 million and non-GAAP EPS of $0.06, the Q4 non-GAAP EPS guide implies that management expects better earnings even if revenue starts at $14 million, which puts pressure on mix, expense control, and the absence of tariff-related impact in the coming quarter. The upside case is straightforward: Q4 revenue above the Q3 level with product revenue doing the work would validate that the $12.7 million product base is expanding as LEO satellite parts move from engineering samples to full production. The downside case is equally clear: if Q4 revenue is near $14 million and EPS still needs below-the-line support, the $0.06 Q3 EPS becomes evidence of earnings fragility rather than a bridge to sustainable profit.

What to watch next quarter is therefore specific and falsifiable. First, Q4 FY2025 revenue must land meaningfully within or above the $14 million to $15 million guide, with particular attention to whether it exceeds Q3’s $14.1 million and approaches the quarterly-history Q4 FY2025 level of $14.8 million. Second, gross margin needs to hold around the Q3 FY2025 level of 51.3% or improve toward the Q1 FY2026 quarterly-history level of 52.7%, because another print near 50.8% would argue that the product ramp is not yet improving unit economics. Third, listen for whether PERSYST EM064LX HR and EM128LX HR moved from engineering samples to full production in the fourth quarter 2025, because that is the cleanest confirmation of the LEO satellite thesis. Fourth, separate IBM FCM4 commentary from Lucid Motors Gravity SUV commentary: IBM should remain consistent with the prior quarter, while Lucid Motors should show volume increase language if the automotive ramp is contributing. Finally, track other income against the $1.2 million recognized in Q3 and $8.5 million to date from the $14.6 million DoD contractor award; the bull case strengthens only if EPS improvement comes with product revenue and gross margin, not another quarter where non-operating income masks operating expense of $8.8 million.