AMD smashes Q3: $13.38B revenue, data-center sales nearly double, 2027 growth over 80%
AMD reported record third-quarter revenue of $13.38 billion, up 45% from a year earlier, as data-center sales nearly doubled — then guided to about $17 billion for the fourth quarter and more than 80% growth in 2027.

The News
AMD reported the biggest quarter in its history on November 3. Revenue reached $13.38 billion, up 45% from a year earlier and above the roughly $13.09 billion analysts expected. Adjusted earnings were $2.09 per share, against roughly $1.95 expected. Shares rose 3.65% in the regular session to $483.90, then climbed a further roughly 7 to 8% in extended trading after the results.
The engine was the data-center division, which nearly doubled. Data-center revenue hit a record $8.55 billion, up 97% from a year earlier and above the roughly $8.20 billion expected. It now makes up 64% of AMD's sales, against 47% a year ago, driven by EPYC server processors and the first Instinct MI450 accelerators shipping in Helios rack systems. AMD says Anthropic, Meta, Microsoft, OpenAI and Oracle are running workloads on those first Helios racks.
The forecast was even stronger than the quarter. AMD guided fourth-quarter revenue to about $17 billion, plus or minus 3%, above the roughly $16.23 billion analysts expected. Chief executive Lisa Su told analysts she expects revenue to grow more than 80% in 2027, with data-center products making up about 80% of sales. There were soft spots, disclosed rather than hidden: the client and gaming division fell 8% to $3.72 billion, missing the roughly $3.79 billion expected; fourth-quarter gross margin is guided to about 55%, down from 56%; and Su flagged a shortage of HBM memory constraining shipments. GAAP earnings were $1.46 per share, including a roughly $1.1 billion non-cash charge for revaluing warrants.
AMD's data-center sales nearly doubled to $8.55B — and it guided 2027 growth above 80%.
Why It Matters
This matters because the AI chip market has had one dominant supplier, Nvidia, and customers have wanted a credible second source for years. A second source means more negotiating power for buyers, more total supply of AI systems, and less risk that one company's production problems stall everyone's plans. AMD's quarter says that second source now has volume: data-center sales nearly doubling, with named customers running its newest systems, is not a promise but a shipment record.
It matters for memory makers, too. Every AI accelerator system needs large amounts of fast memory, and AMD explicitly flagged an HBM shortage holding back shipments. That is a problem for AMD in the short term, but it confirms the demand signal memory investors have been trading on: the bottleneck is supply, not appetite. Micron and SK hynix, named here as read-throughs, sell into exactly this chain.
For ordinary readers, the simplest translation is this: the computers that run AI are being bought at a pace nobody planned for a year ago, and AMD now expects its own sales to grow more than 80% next year. If that forecast holds, 2027 will be the year the AI chip market stops being a one-company story. If it misses, the same forecast will be the reason the stock gives back these gains.
Analyst's Take
The verdict is Bullish, and this time the forecast matters more than the beat. A 45% revenue jump with data-center sales nearly doubling would be a strong quarter on its own. Guiding to about $17 billion for the fourth quarter and more than 80% growth for 2027 turns it into a claim that AMD is no longer a challenger nibbling at the AI chip market, but a second major supplier with a full rack-scale system that the biggest AI builders are already running.
The caveats are real and worth keeping in view. A gross margin guided down to about 55%, an HBM shortage limiting shipments, and a weaker client and gaming division show that this growth is not free: AMD is selling more lower-margin rack systems, and it cannot ship everything customers want. But those are constraints on a boom, not signs of one fading. For Nvidia, the read-through is competition with real volume behind it; for Micron and SK hynix, the read-through is more demand for the memory that goes into these systems. The next test is simple: whether the fourth quarter lands near $17 billion.
By The Numbers
FAQ
What are data-center chips, in plain terms?
They are the processors inside the giant warehouses of computers that run AI and cloud services. Companies rent computing power from those warehouses, so when AI demand rises, the chipmakers supplying them sell more. AMD's data-center division sells both server processors (EPYC) and AI accelerators (Instinct).
What is guidance, and why does it move the stock?
Guidance is a company's own forecast for the next quarter. Investors compare it with what analysts already expected. AMD guided to about $17 billion for the fourth quarter, above the roughly $16.23 billion analysts expected, so the forecast itself was good news even before any sales happen.
What is the difference between GAAP and adjusted earnings?
GAAP earnings follow strict accounting rules and include one-off charges. Adjusted earnings strip those out to show the underlying business. AMD's GAAP earnings were $1.46 per share, including a roughly $1.1 billion non-cash charge for revaluing warrants — paper instruments tied to the share price — while adjusted earnings were $2.09 per share, above the roughly $1.95 expected.
What is HBM, and why does a shortage matter?
HBM, or high-bandwidth memory, is a special kind of fast memory stacked next to AI chips so they can feed data quickly. If HBM is scarce, AMD cannot ship as many AI systems as customers want, even when orders are strong. AMD flagged an HBM shortage as a constraint on shipments.
What is a Helios rack?
Helios is AMD's rack-scale system: a whole cabinet of servers built to work as one AI machine, pairing Instinct MI450 accelerators with EPYC processors. AMD says Anthropic, Meta, Microsoft, OpenAI and Oracle are running workloads on its first Helios racks.
👁
Source: AMD via Reuters / SEC Form 8-K
AI Agent Update