Advanced Energy Is Three Years Early to Its Own 2030 Model — and 44% Below Its High
A record Q3, a 40% growth year, roughly 30% for 2027. The stock has been cut nearly in half anyway.
AEIS · Earnings Call · 2026-11-02
The quarter that broke the model
Advanced Energy's third quarter was not an incremental beat. Revenue of $696 million rose 21% sequentially and more than 50% year-over-year; non-GAAP EPS of $3.65 more than doubled; operating margin reached 24.4%, up 760 basis points. Gross margin of 42.1% was 300 basis points better than a year ago. Q4 guidance is $735 million ±$25 million and EPS of $3.90 ±$0.30. Full-year 2026 revenue growth was raised to "around 40%" from a prior low-to-mid-30% target.
Then CFO Paul Oldham dropped the line that should have moved the stock:
Our Q4 guidance puts us on a run rate to exceed our 2030 target model, 3 years ahead of our original plan.
Translation: the long-range plan shown at the 2024 analyst day is already stale in the best possible way. The 2027 setup was raised too:
For 2027, we think that growth will continue to be robust. I think last quarter, we said we expect it to be over 20%. Directionally, we think that's probably around 30% at this point.
That is a step up from the prior call, when the CFO would only say “We could see over 20% growth next year as well.” — Paul Oldham, Executive Vice President and CFO · 2026-08-03
What actually changed: delivery, not promise
For two years the AEIS story was a pipeline story — design wins at leading-edge nodes, a second wave of data center customers, Thailand. This quarter it became a delivery story, and the keyword set reflects the shift. Factory execution and increasing customer demand now top the list; data center customers and leading edge wafer capacity sit right behind. The old guard — "second wave," "ramp to volume," "piece part inventory," "outgrow WFE" — fell out of the top-30 entirely.
That is not cosmetic. Thailand is real: 500,000 square feet on a single floor, products already coming off the line for two data center customers, with two top semiconductor customers expected to qualify within 2-3 months. Steve Kelley says the factory will take total revenue-generating capacity to "over $5 billion by the end of 2028" and give the company "capacity which will get us through 2029 easily."
The data center business is running hot enough to require expedites: “we're actually having to air ship some of our power supply units because we can't keep enough in stock at the customer hub.” — Stephen Kelley, President and CEO · 2026-11-02
New product vocabulary also appeared: the Power Distribution Board for 800-volt architectures (98% efficiency, 350 watts per cubic inch), and a CPU opportunity — Kelley's point being that any reason to build another rack is good for a power supply maker. He also quantified the semiconductor new-product thrust: “we believe that more than 10% of our semi revenue will come from new products” — Stephen Kelley, President and CEO · 2026-11-02 in 2027.
But two things were quietly dialed back. First, the hyperscale count narrowed. Last quarter: “we have 4 hyperscale customers, but some of those are bigger than others.” — Stephen Kelley, President and CEO · 2026-08-03 This quarter: “we have 2 significant customers in hyperscale. We haven't given out that number before.” — Stephen Kelley, President and CEO · 2026-11-02 Second, the second wave is now explicitly outside the plan: “we have a relatively small amount of revenue in our forecast for the second-wave customers... we're not dependent on second-wave revenue to meet our outlook for '27.” — Stephen Kelley, President and CEO · 2026-11-02 That is conservatism, but it also removes the optionality that made last quarter's second wave narrative so exciting.
The market's rebuttal
None of it helped the stock. AEIS is down 37.6% in 90 trading days and 44.2% from its May 1 peak of $389.05, leaving a market cap of $8.7 billion. This is not a company-specific verdict; it is a sector de-rating. The global tape shows AI data center down 22% over 90 days with 52 declining tickers, "data center capacity" down 45%, "traditional data center" down 46%, "AI cloud services" down 46%. AEIS is simply high-beta to that complex.
There is a real debate underneath. On trailing numbers the stock still looks expensive — 7.3x revenue and 68x net income. But annualize the Q4 EPS guide of $3.90 and you get roughly $15.60; at about $212 per share that is 13-14x. The market is either paying up for a compounder or pricing a peak. The drawdown says peak.
The cash flow is the bear's best argument. FCF less SBC was just $14 million in Q2 against $50 million of capex — a 2.5% free-cash-flow margin and a price-to-FCF of 824x. Inventory sits at 144 days, deliberately, to protect against component shortages. Effective net cash was roughly breakeven in Q2 before the convertible redemption flipped it to $159 million in Q3 — a genuine improvement, but the balance sheet is being used as a buffer, not a distributor.
Supply is the other honest caveat. Kelley flagged constraints in "isolated components, particularly ICs... PCBs and also specialty ceramics," and warned that memory and processor allocation could slow second-wave ramps. Interestingly, the tape is punishing memory-adjacent names — "high bandwidth memory" is down 29% over 90 days with 15 decliners — even as AEIS describes a sold-out memory market. One of those views is wrong.
What to watch
Three markers over the next two quarters. First, Thailand semiconductor qualifications: two top semi customers in 2-3 months would validate the capacity story and de-risk 2028. Second, the 800-volt ramp — still late 2027 into 2028 — and whether the Power Distribution Board converts interest into design wins. Third, whether second-wave revenue starts getting into guidance; right now management is sandbagging it.
And watch the buyback. In October AEIS repurchased about $30 million of stock at around $241 per share — above where it trades now. Management is putting capital behind the idea that this is a base, not a peak. The next two quarters will decide who is right.