Revenue Up 46%, Share Count Up 6.7%: Doing the Math on AIP

The FY2026Q2 report, released after the close on August 6, was close to flawless on the top line. The bottom line was another story, and the market voted with its feet that same evening.

The headline numbers:

String the last five quarters together and you see a revenue line getting steeper every quarter, while the operating loss refuses to shrink:

Growth climbing from 13% to 46% is genuine acceleration. Management blamed the wider Q2 loss on items it described as one-offs: the rising stock price meant an extra $1.7 million in employer taxes when French employees' restricted stock units (RSUs) vested, sales commissions rose on big deals, and $2.2 million of costs tied to the Cycuity acquisition landed in the GAAP numbers (Q2 earnings call). The French tax item really is tied to the share price and should fade if the stock cools. Commissions are a different matter: they follow bookings, so as long as big deals keep coming, that cost stays.

Full-year guidance gives us a hurdle we can actually do arithmetic on. FY2026 revenue guidance was raised from $91 million to $95 million (set at Q1) to $95 million to $98 million. The $96.5 million midpoint sits 2.0% above the $94.6 million consensus of 6 analysts before the report. Q3 guidance of $24 million to $25 million has a $24.5 million midpoint, 4.3% above the $23.5 million consensus (company 8-K and pre-report consensus snapshot, both 2026-08-06). The first half already delivered $47.1 million. If Q3 lands at the midpoint, Q4 needs only $24.9 million to hit the full-year midpoint, just 1.6% sequential growth. The revenue target itself is not a stretch.

Profit is the stretch. Full-year non-GAAP operating loss guidance was loosened from $4.5 million to $8.5 million to a range of $7 million to $10 million. The first half already lost $7.1 million, and Q3 guidance calls for another $1 million to $3 million loss. Work backward and Q4 has to produce roughly $0 to $1.1 million of non-GAAP operating profit, about $0.6 million at the midpoints. So the single number to watch in the November report is whether management's Q4 guidance still implies positive non-GAAP operating profit. The market already takes revenue beats for granted. If breakeven slips again, the acceleration story gets marked down.

Demand is the strongest part of this report. Over the trailing twelve months, customer design starts (a customer formally committing to use Arteris IP in a new chip design) rose 21%, and the CEO said on the call that most new projects involve AI or high-performance computing (HPC). One hyperscale cloud provider has made Arteris its standard interconnect IP for next-generation data center chips. A large U.S. ASIC design services firm is using FlexGen SmartNoC in multi-die chips it builds for hyperscalers. And Li Auto's in-house autonomous driving chip now ships in the L9S and has started paying royalties (Q2 earnings call). Management put it bluntly:

Positive Q2 free cash flow of $8.6 million makes it look like the company now pays its own way. But most of the extra cash that showed up on the balance sheet that quarter came from the stock market.

How heavy is the dilution for existing holders? Shares outstanding rose from 46.0 million at the end of March to 49.1 million at the end of June, a 6.7% expansion in one quarter (company 8-K, 2026-05 and 2026-08-06). Based on $72.5 million and the low end of the disclosed average price, the ATM accounts for at most about 2.07 million shares, roughly 4.5% of the starting share count. The rest came from employee equity vesting and share-based payments tied to the Cycuity deal.

The timing was smart. Selling stock near a high means swapping expensive paper for cash that never has to be repaid, and the company had only $11.7 million in cash and equivalents at the end of March, so this raise thickened the cushion dramatically. Flip it around, though: why would a company that just said FCF will turn positive this year sell more than $70 million of stock in a single quarter? The reasonable explanation is building a war chest for acquisitions and custom development for AI customers. Management has not ruled out further deals, which means more than $100 million of shelf capacity still hangs over the stock. Every leg up in the share price could bring new supply.

Earnings quality needs to be read in two layers. The gap between GAAP and non-GAAP operating loss was $9.3 million: $6.3 million of stock-based compensation (SBC), equal to 26% of quarterly revenue, $2.2 million of acquisition costs, and the rest mostly amortization of intangibles (company 8-K, 2026-08-06). SBC at a quarter of revenue is not outrageous for a small-cap IP company, but it does mean the "near breakeven" non-GAAP figure pays a meaningful slice of employee compensation through a growing share count. Q2's 6.7% expansion is part of that bill. Cash conversion, by contrast, is a bright spot. IP licensors usually collect cash before booking revenue, so Q2 operating cash flow of $9.2 million was far better than the GAAP net loss, helped by seasonal inflows of prepayments and deferred revenue. That rhythm makes quarterly FCF swing wildly, so Q1's negative and Q2's positive should be read together.

Arteris sells NoC IP (network-on-chip, the data network that links the different blocks inside a chip) and SoC integration software. It has no wafers and no factories, so it has almost no upstream supply chain. Its real "raw material" is engineers who understand on-chip interconnect design. Q2 R&D was $16.8 million, up 38% and equal to 70% of revenue (10-Q, 2026-08-06). The French tax charge is a reminder that a large share of the R&D team sits in France, and labor costs move with equity compensation: the better the stock does, the pricier the people. That is the real bottleneck in this business.

The customer base is more spread out than many assume. The company says no single customer exceeded 10% of revenue in H1 2026 (Q2 earnings call). Enterprise computing averaged 29% of ACV plus royalties over the past four quarters, with the rest spread across automotive, communications, consumer electronics and industrial. Named or described data center customers include a hyperscaler that has standardized on its IP (unnamed), a U.S. design services firm building ASICs for hyperscalers (unnamed), and AMD, which announced in August 2025 that it is using FlexGen in AI chiplets. On the automotive side there are Li Auto and SiEngine. Arm also expanded its license to Arteris's hardware security verification technology in Q2 (company 8-K, 2026-08-06; AMD partnership per company announcement, 2025-08).

It is worth being clear about how firm the contract figures are. ACV plus royalties of $99.5 million is annualized contract value plus the last four quarters of royalties, a measure of "how much has been signed." RPO of $135 million is the firmer number: signed obligations not yet recognized, more than half due within 12 months. Both are well above current revenue, but getting from a signed deal to meaningful royalties means going through a full cycle of chip design, tape-out and volume production, usually one to two years or longer. Royalties have already shown a fragile side: Q2 royalties slowed from the prior quarter, which the CFO attributed to one customer's logistics and supply chain problems pushing shipments back a quarter (Q2 earnings call).

The weakest link is royalty visibility. License revenue depends on deal signings, a pace Arteris negotiates with customers. Royalties depend on how well customers' chips sell, and Arteris has zero control over the production ramp of hyperscaler custom ASICs or automotive self-driving chips. Management sticks to a long-term royalty compound growth rate in the "high thirties to low forties," but delivering on that depends on a handful of AI ASICs actually ramping in volume in 2027.

Few listed companies are directly comparable in semiconductor IP. Arm is the most direct reference point, while AI interconnect and custom chip companies in the same group show how hot Arteris's end markets are.

The direction lines up: the acceleration in Arteris's design starts and royalties roughly matches the strong growth at downstream AI chipmakers. The difference is that Arteris sits at the very front of the chain and gets paid when projects start. The downstream shipment peak will only flow back into its numbers as royalties a year or two later.

How does the market price results like these? As of October 6 this season, the only comparable peer that has reported is PRGS. It reported after the close on September 30, beat EPS estimates by 19.0%, and still fell 8.5% the next day, close to the ±10.2% move options had implied (peer earnings reaction data, as of 2026-10-06). Beating and still falling suggests the market had already priced in good news for richly valued growth stocks and was focused on the margin and the outlook. This reading has limited value, though. PRGS is a software company, and over the last 8 quarters its earnings reaction moved in the same direction as Arteris's only 4 times, a correlation of r=0.20, essentially none. It is not a reliable guide to Arteris's next earnings move.

Arteris's own history is more useful. Over the last 8 quarters, the median earnings move implied by options was ±15.4%, while the median actual move was only 9.8%, and the actual move exceeded the implied one just 3 times (as of 2026-10-06). The options market habitually overestimates this stock's earnings-day swings, and August's roughly 5% after-hours drop fit that pattern.

The company is still losing money on a GAAP basis, so price-to-earnings (P/E) and EV/EBITDA are useless here. The right tools are EV/Sales (enterprise value divided by revenue) alongside forward order measures like RPO and ACV, then checking the multiple against the growth rate.

What does 11x forward EV/Sales mean? For an IP company guiding to roughly 37% revenue growth in 2026 at the midpoint, with gross margins above 85%, that multiple is moderately high within AI semiconductors: below Arm, but well above the historical average for small-cap EDA and IP names. Divide by growth and each percentage point of growth costs about 0.3x EV/Sales, which on its own is not extreme.

At this multiple, the market has already paid in advance for two things. First, that 2026's 37% growth carries into 2027 without a sharp slowdown. Second, that once Q4 non-GAAP operating profit turns positive, margins keep expanding as royalties ramp. The first has RPO and design starts behind it. The second, so far, rests on a single management phrase: "as early as the fourth quarter." The valuation isn't outrageous, but almost all of the margin of safety rides on the royalty timeline. If the Q4 breakeven calculated in Chapter 1 gets pushed out, the multiple will adjust before revenue does.

Arteris's cycle is not the four-year capacity-and-pricing cycle of memory or foundries. It follows the chip design-start cycle: when a new generation of architecture arrives, lots of customers launch projects and sign licenses at once. A year or two later, those chips go into production and royalties ramp. Once the architecture settles and new projects thin out, license growth slows first and royalties peak later.

Three readings locate where we are now:

Our read: the AI-driven design-start cycle is in the middle of its upswing. Signings have accelerated for several straight quarters, royalties are just starting to follow, and non-AI lines like automotive and consumer electronics are relatively flat. Two leading signals to track: first, year-over-year growth in ACV plus royalties. A drop from above 40% to below 30% would mean the peak in new projects has passed. Second, hyperscaler capex guidance. A broad cut would hit new ASIC projects first, and that would show up in Arteris's bookings two to three quarters later.

This chapter covers week-level signals for timing entries and exits over the next few weeks. Readers holding for a year or more can skip it.

Arteris isn't part of any sector basket with money-flow tracking, and it isn't in any of the current rotation themes, so there's no money-flow reading for its specific niche. We can only judge its backdrop from the broad market and investment-style level (all data as of 2026-10-06).

The broad backdrop is neutral, tilting slightly toward risk-taking. The overall market reading is neutral, with a composite score of 0 (on a scale of -100 to +100). SPY closed at 781.10, about 2.00% above its 20-day moving average of 765.81, and the gap between the 20-day and 50-day averages is only +0.07%. The cross-asset rotation signal reads risk_on: money is moving from consumer staples (XLP) into technology (XLK) with strength 4.86, and from energy (XLE) into XLK with strength 3.10. Two of the three sub-sectors with the strongest inflows are semiconductor equipment: metrology and inspection, represented by KLAC (flow strength 77), and power and vacuum control, represented by MKSI (75). The third is liquid biopsy, represented by GRAL (72).

For a small cap like Arteris, though, two readings pull the other way. At the style level, outflows are concentrated in Russell 2000 small caps (IWM), at 14% of the total. On macro liquidity, net liquidity is $5.78 trillion, up just 0.2% week over week, and the 10-year Treasury yield is 5.28%, putting liquidity conditions on the cautious side (FRED, 2026-10-06). In trader terms: money is going into tech, but into large caps and the leaders of the equipment and computing supply chain. Small-cap growth stocks are being sold down, and high rates are weighing on the valuations of long-duration assets.

The stock's own chart tells the same story. It sits slightly below both its 50-day and 200-day moving averages, and the two lines are nearly on top of each other, meaning the past few months have been a directionless sideways drift. The stock hasn't run ahead of the money, and the money hasn't come back to push it.

Long term (quarters to years): worth following. Revenue growth has accelerated for four straight quarters to 46%. RPO of $135 million gives visibility into more than half of next year's revenue. Design starts and standardization at a hyperscaler show Arteris has won a real position in AI chiplet interconnect, and roughly 11x forward EV/Sales for growth near 40% isn't wildly expensive. The costs you have to accept are just as clear: SBC at a quarter of revenue, a 6.7% share count expansion in one quarter, more than $100 million of shelf capacity left for new share sales, and a royalty pace entirely at the mercy of customer shipments. The way to hold it is to follow the checkpoints. In each quarterly report, look at just two things: whether ACV plus royalties growth is still above 35%, and whether Q4 non-GAAP operating profit turns positive on schedule. If both hold, keep holding. If breakeven slips, move it from the holdings list back to the watch list.

Short term (days to weeks): don't chase. Wait for money to return to small caps and for the stock to reclaim its moving averages. It isn't part of any money-flow theme, small-cap stocks are in a selling phase, a 5.28% 10-year Treasury yield is weighing on growth valuations, and the stock is drifting sideways below two converged moving averages. Right now there's neither a money-flow tailwind nor a favorable entry point. The next earnings report is only a month away in early November, and since options have historically overestimated its earnings moves, positioning ahead of it isn't a good bet.

The sections below turn these two views into specific conditions and price levels.