Guidance 11% Above Consensus, Yet the Stock Sits Below Its Averages

Monolithic Power Systems (MPS, ticker MPWR) designs the chips that manage and deliver electricity inside data center servers, cars, networking gear and other electronics. The headline results:

The growth this quarter was strikingly concentrated. Of the $465.2 million in added revenue compared with a year ago, Enterprise Data supplied $315.2 million and Communications supplied $107.5 million. Together, that is nine-tenths of the increase. The company's "diversified growth" description holds up as a long-term strategy, but today's engines are still data centers and optical communications.

The quarter-by-quarter trend says more than any single quarter (company 8-Ks for each quarter):

Revenue climbed by more than 60% in a year while gross margin moved just 0.2 percentage points, and that nearly flat line is itself the message. Management said on the call that pricing is "really not a significant contributor" and that "we definitely absorb some of the cost increase" (earnings call, 2026-10-29). In a year when peers have been raising prices one after another, MPS is growing mainly through shipment volume and through modules that raise the dollar value of each sale. That kind of growth tends to wear better than growth built on price hikes. It also means margins have less room to expand than many bulls imagine. Management calls 55% to 60% the "sweet spot" and says it does not want to push margins up one quarter only to give them back the next.

The full-year math can be pinned down precisely. Revenue for the first three quarters totals $2.9872 billion. Add the fourth-quarter guidance midpoint and the full year comes to about $4.34 billion, up about 55% from roughly $2.79 billion in 2025, inside the 50% to 60% range management described. The number that really matters is a different one. Management raised its floor for 2026 Enterprise Data growth from 130% to 150%. That business was $701.8 million in 2025, so the full-year floor is about $1.755 billion. The first three quarters delivered $1.1501 billion, which means the fourth quarter needs to reach at least about $604 million, up about 19.3% from the prior quarter.

That $604 million is the number most worth watching in the next quarterly report. At the guidance midpoint, it would absorb about two-thirds of the $147.6 million in sequential revenue growth. The old 130% floor worked out to only about $464 million, below what the third quarter already delivered, so it had long since stopped meaning anything. The upgrade reads more like an acknowledgment of reality than a fresh promise.

The evidence on demand is solid. The press release says "We saw strong demand across our existing customers which we expect to continue into 2027." Founder and CEO Michael Hsing used the phrase "explosive growth period in the next few years" on the call, and channel inventory sits at the "lean end of our range." But when the CFO was pressed on 2027 Enterprise Data growth, the only reply was "it's a little early" (company 8-K and earnings call, 2026-10-29). The CEO is talking up the boom and the CFO is not giving a number, and part of the stock's hesitation comes from that gap.

This company is not short of money. It holds $1.0912 billion in cash and short-term investments, carries no borrowings on its balance sheet (company 8-K, as of 2026-09-30), and faces no pressure to raise funds or issue new shares. The question worth digging into lies between the income statement and the cash flow statement. The sudden widening of the gap between GAAP and non-GAAP profit in the third quarter is the part of this report most easily glossed over.

Two takeaways. First, neither the press release nor the call records we could find break down why SBC doubled, so the cause is unknown. If it includes extra accruals on performance-linked stock units as growth runs ahead, at least part of it will rise and fall with results, and the higher fourth-quarter guidance shows this is not one-off noise. Anyone valuing the stock on non-GAAP EPS is erasing more than $100 million a quarter of real equity cost. Second, GAAP operating profit grew 17.7% from the prior quarter while net income grew only 5.4%. The difference falls mainly on the income tax line, and the specific cause will have to wait for the 10-Q.

On the cash side, the company bought back $453.4 million of stock in the third quarter, equal to 156% of that quarter's operating cash flow. Add the quarterly dividend of $2 per share (company announcement, 2026-09-10; about $100 million based on 48.8 million shares), and cash on hand fell $322.6 million from the prior quarter. On top of that, the board added another $500 million authorization in October, bringing the total back to $1 billion. Don't be dazzled by that figure. Fourth-quarter diluted share count guidance has a midpoint of 49.1 million, only about 0.2% below the third quarter's 49.2 million. SBC is offsetting the buybacks, so those hundreds of millions of dollars are mostly buying "no dilution" rather than a meaningfully smaller share count.

Free cash flow (FCF), the cash left after capital spending, also deserves a closer look. Full-year 2025 FCF was $838.2 million, an FCF margin of about 30%. In Q2 2026, FCF was $147.3 million, with the margin down to about 15% (company announcements). Third-quarter capex and FCF were not in the press release and will have to wait for the 10-Q. The CFO said "I don't anticipate at this point seeing a large increase in capital intensity," yet net fixed assets reached $821.6 million at the end of the third quarter, up from $627.7 million at the start of the year. Tying up working capital is a normal price of rapid growth, but it should be said plainly that this AI upswing is producing less cash than the income statement suggests.

There is also a governance item on the ledger. In its February 2026 10-K, the company restated 2024 deferred income taxes by $194.6 million and disclosed a material weakness in internal controls over financial reporting related to deferred tax accounting, with remediation and testing expected to be completed during 2026 (FY2025 10-K). This has no direct effect on operations, but it does explain why the $32.60 EPS for 2024 cannot be used as a baseline. Whether the next 10-K declares the issue fixed is a hard checkpoint.

Upstream, MPS is a classic fabless company, meaning it designs chips but outsources manufacturing. Its wafers are produced by multiple foundries in mainland China, Taiwan and South Korea (company 2026 GRI report), and it also has a significant business presence in Taiwan (FY2025 10-K). Management says capacity at the end of the third quarter could support more than $6 billion in annualized revenue, reaching $8 billion in the second half of 2027 and continuing toward $10 billion (company 8-K, 2026-10-29).

Annualize the fourth-quarter guidance midpoint and you get $5.4 billion, already nine-tenths of existing capacity. Picture a restaurant with nine-tenths of its seats booked and a new branch that won't open until the second half of next year. In the CFO's words, "capacity stays ahead of the business," so capacity is not a bottleneck today. But room to grow in the first half of 2027 really is tied to the pace of expansion.

There are three real constraints. Wafer costs are rising, and the company is choosing to absorb part of that increase and share the rest with customers. The memory shortage has spread to laptops, producing what the CFO called "a tale of 2 cities" inside Storage and Computing: power management chips for DDR5 memory are growing while laptop demand is soft, leaving this market roughly flat to up a few points for the full year. The third constraint sits downstream.

The customer picture has layers:

That third point is the most fragile link. The distributor figures make concentration look like a sales-channel issue, but the real risk in this business is losing share on a single platform. During the Blackwell generation, reports said some share went to Infineon and Renesas (media reports). "No purchase commitments" means that visibility on demand rests on design wins and ordering patterns, with no contractual backstop.

The company's hedge is to widen its customer base. The CFO pointed to rapid growth among ASIC customers (companies using custom AI chips), who mainly buy modules with higher value per chip, under a strategy of being "in every socket." The press release also mentions using in-house AI design tools to reach more customers. The direction is right, but with Enterprise Data now 42.1% of third-quarter revenue, public numbers can't yet tell us how much platform risk this diversification offsets. Management also declined to comment when an analyst asked about competition from ADI's acquisition of Empower (earnings call, 2026-10-29), which is another thing to watch.

The fair comparison is with other companies selling data center power products. Memory and chip-packaging names are many times larger or smaller and offer little useful comparison, so we limit this to four companies:

Three things stand out. On growth, MPS leads this group and is no longer small. Applying TI's 13% share implies about $800 million a quarter (calculated from the disclosed ratio), still larger in absolute terms, but growing at only six-tenths of the MPS rate. TI's overall fourth-quarter guidance is also flat, while MPS is still growing by double digits from the prior quarter. On profitability, MPS has no edge. TI's gross margin is about 8.5 percentage points higher, partly because TI owns its factories, and that suggests the MPS advantage lies in the speed of winning share rather than in pricing power. Vicor's 68.3% gross margin is visibly lifted by licensing revenue and is not directly comparable.

On direction there is no disagreement. TXN (2026-10-21), NXPI (2026-10-27), and STM and ALGM (2026-10-29) all cited stronger demand as a driver this quarter, which shows that analog and power semiconductors more broadly are rising, not just AI power. There is a flip side: when peers are all thriving, MPS looks less unique, and the premium the market pays for it has to be sustained by a continuing gap in growth.

As for how the market is pricing similar report cards, there is currently no disclosed data this quarter on comparable peers' earnings-day stock reactions or options-implied volatility to compare against, so we leave that blank rather than guess.

The 73.7 times trailing P/E (share price divided by the past year's earnings) is not much use here. Its denominator is the past year's GAAP profit of a company growing at more than 60%, muddied by doubled SBC and an elevated tax rate. Forward non-GAAP multiples, meanwhile, erase the cost of equity pay. So we use both yardsticks and read the answer as a range.

The 23-times gap between 32 times and 55 times is the gap created by SBC and taxes, and which one is closer to the truth depends on how you view SBC. We lean toward treating it as a real cost. For pricing, we center on a non-GAAP full-year multiple of around 40 times, rather than taking comfort in the annualized 32 times.

So what has this valuation already paid for? Consider one scenario. If the $8 billion of capacity really is filled in the second half of 2027, then a 40% non-GAAP operating margin and a 15% tax rate would produce annualized EPS of about $56, putting today's market value at about 22 times. That shows the market has not fully priced in the full-capacity case, which is why the PEG and forward multiples still appear to leave room. But the flip side is that the stock has already paid in advance for a sizable stretch of the road from $5.4 billion to $8 billion. If the central expectation for 2027 growth falls short of what the market assumes, or if share on a major platform is lost, multiples and earnings could come under pressure together. With the stock about 29% off its high while fundamentals keep getting revised up, it is indeed much cheaper than six months ago, but it still does not offer a margin of safety.

Our view: this analog and power semiconductor cycle is in the middle of its upswing. AI power is running ahead of the overall cycle, but still has some distance to go before any signs of a peak. Three sets of readings support this.

We don't call this the late stage because inventory days are still falling and the sales channel shows no buildup. We don't call it the early stage either, because price hikes are already on their third round and new capacity is expected to arrive in phases from 2027 to 2028 (Futurum analysis), which happens to coincide with MPS pushing its own capacity to $8 billion.

Two leading signals are worth tracking:

This chapter deals with week-level signals that bear on timing over the next few weeks. Readers holding for a year or longer can skip it.

However good the report, money has not been on the side of semiconductors these past few weeks. As of 2026-10-30, the sector money flow reading for the semiconductor index (SMH) stands at -63 on a scale of -100 to +100, among the top three outflows alongside lithium batteries at -74 and quantum computing at -59. The largest inflows went to IT services (ACN) at 75, chemicals (LIN) at 67 and defense technology (SHLD) at 67. Where the money is going is just as clear: the two strongest rotation pairs that day were out of technology (XLK) into real estate (XLRE), at a strength of 4.01, and out of XLK into materials (XLB), at a strength of 3.90.

The wider backdrop is just as chilly. As of 2026-10-30, the market regime reading is -28, in bear market territory, SPY sits 1.91% below its 20-day moving average, and the broad asset signal is risk-off. On the macro side, net liquidity is $5.75 trillion, down 1.0% from the prior week, and the 10-year Treasury yield is 4.98% (FRED, 2026-10-30). Healthcare (XLV) is the direction that has actually formed a trend, helped by earnings from Eli Lilly and Merck, while consumer discretionary and industrials are still building. The power chip niche has no separate money flow reading and isn't among any of the directions now drawing inflows, so semiconductors as a whole is the only available guide.

In trader's terms, crowded positions in technology and semiconductors are being cut, and the money is heading to defensive and cheaper cyclical stocks. A high-octane name like MPWR, with a 73.7 times P/E and a beta (sensitivity to market swings) of about 1.68 (Investing.com), is easily treated as an ATM. The share price fits this picture: down 1.02% on earnings day, down another 2.63% after hours, then up 1.1% in the latest quote. It sits about 5.1% below its 50-day moving average and about 7.9% below its 200-day moving average, down 21% over the past six months yet still up 33% year to date (Investing.com and live quote). This time the stock has fallen behind the fundamentals, rather than running ahead of the money. In that situation, a beat can cushion a decline but is not enough on its own to reverse the trend. That has to wait for the money to come back.

Long term (quarters to years): worth following. Enterprise Data more than doubled over three quarters, Communications more than doubled from a year earlier, gross margin is stable, and the inventory picture is actually improving. For 2027 there is $8 billion of capacity plus several new growth lines, including ASICs, 1.6T optical modules, 48V and data converters. The fundamentals pass. At about 41.5 times full-year non-GAAP earnings and about 55 times annualized GAAP earnings, the valuation is not cheap but is in keeping with the growth rate. This is a trade of valuation discomfort for growth visibility. There are four costs to accept: real profit shrinkage from the SBC jump; share shifts on a single platform with no purchase commitments; multiple compression if 2027 growth slows as capacity fills; and internal control remediation that isn't yet finished. Holding should be driven by checkpoints: fourth-quarter Enterprise Data reaching about $604 million, next-quarter guidance still pointing to sequential growth, and forward inventory days not rising. If all three hold, keep following it. If not, reassess.

Short term (days to weeks): don't chase; wait for the money and for a better price position. A sector reading of -63, risk-off conditions, shrinking net liquidity and a share price stuck beneath two moving averages (all as of 2026-10-30) all say that buying now means stepping in front of a stock still being sold down. The earnings beat provides a cushion on the way down. Only when semiconductors leave the top of the outflow list and the shares reclaim their medium-term average will the short-term odds shift back toward readers.

The next two chapters translate these views into specific conditions, price levels and tracking items.