Fiscal Q2 2027 earnings call, November 5, 2026: record-quality beat, but a softer December guide and a lower margin outlook

Microchip Technology reported September quarter net sales of $1.622 billion, up 9.2% sequentially and 42.2% year over year, and guided the December quarter to a range of down 3% to up 1%. The headline numbers were clean, but the more important news came in the commentary. Management walked back its prior assumption that gross margin could be held near 66% and now says investors should treat 65% as the ceiling, with the cause being a supply chain that cannot keep pace with demand rather than weak demand itself.

Gross margin: management concedes last quarter's assumption was too aggressive

Non-GAAP gross margin was 66.5% in the September quarter, helped by about $53 million of infrequent items that carried a 100% gross margin. These were roughly $21 million from a one-time revaluation of distributor price protection after the August price increase, $22 million of licensing revenue above the expected run rate, and a $10 million benefit from a long-term customer contract. The December guide of 64% to 65% reflects the loss of those items. It also reflects substantial unsupported backlog in high-margin products that suppliers cannot fulfill, and lower internal factory utilization than planned a quarter ago.

CEO Steve Sanghi was direct about the revision. "I think our assessment last quarter that we can keep the gross margin in the 66% range by more high-margin products coming in, the rest of the price increase kicking in and completely offsetting the higher licensing revenue and a couple of other infrequent items, that turned out to be more aggressive than the reality," he said. "We are basically resetting that and saying, think of 65% as a long-term gross margin, and we will work very hard to get there and maintain it, but not any higher than that."

Input costs are the next pressure point. Sanghi said another round of price increases from most major foundries and outsourced assembly and test providers (OSATs) begins in January. Microchip does not plan a second customer price increase so soon after the mid-August one and intends to absorb the supplier increases for now. "We haven't figured that out yet," Sanghi said of whether the company will accept a margin hit or pass the costs on later. Underutilization charges were $26.6 million in the September quarter. CFO Eric Bjornholt said the December quarter will not repeat that improvement, because the company is slowing back-end assembly and test after building finished goods last quarter, while front-end fabs ramp to rebuild depleted die inventory. Further improvement is possible in later quarters, though not in a straight line.

Supply, not demand, is the binding constraint, and AI is the cause

Management described a pivot in September, with customers finishing inventory digestion and shortages appearing as lead times lengthened. Standard-product lead times historically ran 4 to 8 weeks and now range from 6 to 32 weeks. Bookings were very strong, book-to-bill stayed well above 1, and backlog entering December was higher than a quarter earlier. Many orders inside the lead time are going unfilled, and some are being pushed into the March quarter, which Sanghi said has a notably better backlog than December had at the same point.

The most notable claim was that AI data center demand is displacing Microchip's capacity rather than its customers' demand. "Data center has been crowding out capacity for trailing edge nodes," Sanghi said, adding that at OSATs, "AI companies came in and took over the whole factory and they pushed everybody out." He said the displacement hits internally fabricated wafers as well, since many packages are assembled and tested outside, and that "our test capacity has been confiscated by the AI cloud." Remedies such as qualifying alternate OSATs, loading minimum batch sizes and insourcing packaging take about six months, and some displacement emerged only a couple of months ago. Management expects incremental capacity in the second half of the December quarter and more in the June quarter. Sanghi ruled out revisiting a 300-millimeter fab, calling the problem a short-term one.

Customers are not yet responding. Sanghi said OEMs are showing no appetite to build inventory and compared chip makers' requests for longer backlog to "the wolf is coming" warnings that go ignored "until they get in trouble."

Why the December guide is better than it looks

Stripping out the $53 million of infrequent items, Sanghi said December is up about 2.3% sequentially against normal seasonality of down 3% to 5%. About $10 million of that comes from acquisitions, with $4 million from Hailo and $6 million from Vorago, so organic growth is closer to 1.5%. The guide sits roughly 5% to 7% above normal seasonality. Non-GAAP EPS is guided to $0.87 to $0.95, against $0.97 in September and up 107% year over year at the midpoint. Operating expenses are guided at 26.5% to 27% of sales and operating margin at 37% to 38.5%. The long-term model of 65% gross margin, 25% operating expense and 40% operating margin is unchanged.

The data center: small today, and Sanghi calls it a bubble

Data center sales exceeded $250 million in the September quarter, up 55.7% year over year, and are on track for $1 billion in calendar 2026, split about evenly between standard products and the Data Center Solutions unit. Data center was 16.7% of sales excluding licensing. Sanghi said plainly that "data center is in the bubble," and argued that the demand was "explosive" and that its second derivative "has to come down substantially," citing constraints from real estate, power, politics and circular financing.

His argument is that Microchip is insulated on the downside. Roughly half of data center revenue, or about 8% of total sales, comes from broad-based microcontroller, analog, power, Ethernet and USB products where data centers are a small share of shipments. The other half comes from Data Center Solutions, which lost the PCIe Gen5 generation by arriving late and is now rebuilding share. Gen6 switch and retimer design wins rose to 19 at the end of September from 12 at the end of June and stand at 24 today. Production shipments began last quarter, with a sizable shipment this quarter, and Sanghi expects almost all 24 designs to be in production by the start of 2028. He declined to dollarize the wins. "For us, it's all upside," he said, and the upside is "not in our numbers." He conceded that volumes at those wins could come in lower if the build-out slows, since the company cannot always tell whether a customer is shipping into a data center that exists or one a governor may put on hold.

Management now expects a deceleration in AI to help Microchip by freeing capacity. Sanghi said this is the only thing that will resolve the constraints. He also moved Data Center Solutions, led by Brian McCarson, to report directly to him instead of through the networking and connectivity pillar.

Analyst William Stein pushed on whether a bubble unwinding could spill into industrial and aerospace and defense demand and into pricing. Sanghi rejected that, arguing that non-data-center markets are "normal" and healthy and that Microchip should outperform data-center-concentrated peers in a correction. That position rests on a judgment that the broad-based end markets are not being propped up by AI spending, which the company did not demonstrate on the call.

End markets: industrial leads, automotive is above trend

Industrial was 34.6% of sales and grew 50.3% year over year, which Sanghi attributed to a late recovery and design wins across robotics, medical, process control, factories and farming. Aerospace and defense, 16.1% of sales, grew 41%. Sanghi cautioned that growth there will be steady rather than explosive because primes are constrained by floor space and equipment. "It's not a hockey stick because the industrial complex cannot grow at that rate," he said. Automotive grew 19.3% and Sanghi said it was "more than entitlement" given unit growth of 2% to 3%, so investors should not expect more. Communications grew 43.5%, consumer appliances 26.7%, and compute fell 1.6%.

Hailo reshapes the edge AI road map; Vorago is already profitable

Sanghi gave some detail on the Hailo acquisition. Microchip's in-house microcontrollers and microprocessors had no hardware AI acceleration. The internal NPU road map would have delivered a first product in about 2.5 to 3 years and a second in 4 to 4.5 years, and Hailo "accelerates the road map by 5 years." Microchip has canceled both internal NPU products, merged Hailo with the microprocessor unit and is advancing Hailo's road map. Hailo is currently dilutive, with gross margins below the company average, high R&D and only about $4 million of quarterly revenue. Sanghi cited a funnel of multiple hundreds of millions of dollars. Vorago, at about $6 million of quarterly revenue, carries higher-than-average gross margins and small operating expenses.

Balance sheet and capital return

Inventory rose 9 days to 184 days as Microchip received high-value wafers from foundries ahead of expected growth and built finished goods. Distributor inventory was flat at 25 days, at the low end of the historical range. Net debt to adjusted EBITDA fell to 2.36x, adjusted free cash flow was $377.1 million, and net debt fell $31.7 million even after funding Hailo. Capital expenditures are expected at about $100 million for fiscal 2027. Bjornholt said the board wants leverage lowered further after "a pretty big scare during the last downturn," so the dividend stays flat and buybacks or an increase will be evaluated in the next fiscal year.

Sanghi ended the Q&A by saying that even under extremely conservative modeling for fiscal 2028, which begins April 1, he arrives at earnings of "over $4" per share, and that at the current share price he is "tempted to bite myself." That is a promotional remark from a CEO whose gross margin outlook has just been lowered, and investors will want to see the March quarter backlog convert and the supplier cost increases absorbed before giving it weight.