【TMGM Financial Recap】Two Earnings Reports Temporarily Allay AI Infrastructure Concerns as AMD and Supermicro Jump After Hours

A gain of more than 9% in AMD shares initially marked trading after the close on 3 November, before the advance eased to around 6.85%. Supermicro, which also rose, was nearly 10% higher at one point. The earnings released that evening conveyed the same message from two different stages of the AI infrastructure supply chain. For AMD, the business is supplying computing chips; for Supermicro, it is assembling those chips into servers and delivering them to data centres.

The preceding week’s developments give those results added weight. Negative free cash flow for a second consecutive quarter, together with a forecast for a “significant increase” in capital expenditure for 2027, sent Alphabet shares down 10% after hours on 28 October. A 13% decline followed at the next day’s close. On 30 October, fourth-quarter guidance substantially below expectations and an indication that full-year free cash flow would remain negative prompted a 30% one-session fall in Intel. The market is beginning to price in an expanding AI capital expenditure cycle with an increasingly uncertain timetable for returns. While the AMD and Supermicro reports did not establish when those returns might arrive, they addressed the more immediate issue of whether orders for AI infrastructure are holding up.

AMD’s Outlook Takes Precedence as Data Centre Revenue Nearly Doubles and Earnings Beat Forecasts

With third-quarter revenue up 45% year on year at $13.377 billion, against a market forecast of $13.09 billion, AMD surpassed expectations across its results. A two-percentage-point annual improvement took gross margin to 56%. Adjusted earnings per share, meanwhile, reached $2.09, compared with the anticipated $1.95.

Much of the share-price response after hours stemmed from data centres. A 97% year-on-year increase brought divisional revenue to $8.546 billion, lifting its proportion of group revenue to 64% from 47% a year before. Both demand for EPYC server CPUs and increased production of Instinct MI450 series GPUs contributed to the growth. Through the MI450 AI accelerator product line, AMD competes with Nvidia. The reported “record sales” suggest that its position in AI computing can now be measured in sales, rather than solely in a product roadmap.

The greater significance lay in what AMD projected for the fourth quarter, rather than what it reported for the third. Its revenue forecast is approximately $17 billion, plus or minus 3%, putting the midpoint around $700 million above the approximately $16.3 billion average analyst estimate. This implies an increase of around 27% quarter on quarter and approximately 66% year on year. Management is therefore anticipating further acceleration following the third quarter’s near-doubling. Explaining the outlook on the earnings call, Lisa Su referred to the expansion of production capacity for Helios rack-scale AI systems and sixth-generation EPYC “Venice” processors. Her statement was explicit: “data centre AI will grow faster than the company previously expected.”

Elsewhere, performance was less consistent. Gaming revenue contracted 40%, helping to pull combined client and gaming revenue down 8% year on year to $3.722 billion. Higher inventories, accounts receivable and capital expenditure accompanied an approximately 64% annual reduction in free cash flow to $554 million. That was short of the roughly $1.05 billion expected by the market.

Supermicro Eases Concerns as Revenue Grows Nearly 300% and Margins Recover

An approximately 298% year-on-year increase took Supermicro’s revenue to $19.9 billion in the first quarter of fiscal 2027, substantially ahead of the $16.4 billion market forecast. Adjusted earnings per share exceeded the expected $1.23 by approximately 47%, reaching $1.81. For the second quarter, the company forecast revenue of $19.5 billion to $21.5 billion; the market had anticipated approximately $17.1 billion to $18 billion. It also lifted its full-year revenue forecast, replacing a range of $65 billion–$72 billion with $72 billion–$82 billion.

As a business centred on AI server integration and delivery, Supermicro measures activity at a different stage from chip design orders. Its revenue growth directly captures the speed at which cloud providers, AI laboratories and enterprises put computing capacity into data centres. An annual increase of 298% thus signals that the physical build-out of AI infrastructure continues at a rapid rate.

More striking still was the improvement in profitability. First-quarter gross margins reached 12.7% under GAAP and 12.8% on a non-GAAP basis, compared with 9.3%–9.5% a year earlier, and exceeded analysts’ previously pessimistic estimates. The question occupying the market was no longer simply whether Supermicro could increase revenue. It was whether profitability could be maintained against a backdrop of constrained Nvidia chip supplies and higher memory prices. The improvement in gross margin went some way towards easing those doubts.

There are limits to the reassurance these results provide. Legal issues have recently added structural pressure at Supermicro, beyond the fluctuations of the AI cycle. Shares previously fell 33% on the day after news emerged that the US Department of Justice had brought export-related charges against three individuals, one of them a co-founder. To some extent, the post-results rise after hours therefore shows financial fundamentals taking temporary precedence over legal concerns; it does not show that legal risk has gone away. Other issues remain incompletely reflected in the gain, including the company’s change of CFO and the constraints on its business from GPU and processor shortages.

AI Deployment Accelerates While Gains Remain Uneven Across the Supply Chain

The acceleration in chip orders is visible in AMD’s data centre revenue. Increased MI450 production also indicates that the market is absorbing computing capacity from vendors other than Nvidia. At the server stage, Supermicro’s revenue and outlook point to deployment maintaining its pace, with growth of 298% reflecting further expansion in physical data centre capacity. The combined evidence supports a specific conclusion: the “AI capital expenditure anxiety” set off by Intel and Alphabet in late October has not been accompanied by evidence of deteriorating fundamentals in infrastructure orders and deployment, at least.

That conclusion does not, however, equate strong orders and deployment with realised returns. Declining free cash flow remains an issue at AMD. Supermicro’s recovered margins are still around 12%, and free cash flow shortfalls continue at Alphabet and Amazon. Within the AI infrastructure chain, reported figures appear stronger towards the construction end, in chips and servers. Moving towards monetisation, through cloud services’ free cash flow and advertising conversion efficiency, exposes more of the pressure.

For the time being, investors are prepared to pay a premium for continuing acceleration in orders, as the after-hours gains in AMD and Supermicro demonstrate. The test, once the market opens, is whether that premium holds and extends to the whole AI infrastructure chain. Both will depend on investors’ judgement of whether order strength can outweigh uncertainty over the timetable for returns.