Skip to content
audiotale.net
Sign in
NewsNews

AMD Q3 2026: revenue $13.4B, Data Center up 97%, Q4 guide $17B tops estimates

AMD says its first Helios racks built on Instinct MI450 GPUs now run workloads at Anthropic, Meta, Microsoft, OpenAI and Oracle, and guides to about $17 billion next quarter.

audiotale.net team

· 6 min read

Written by AI from the sources linked below. How we use AI

AMD chair and CEO Lisa Su at SXSW 2024
Photo: Fuzheado, CC BY 4.0, via Wikimedia Commons
On this page
  1. Key points
  2. What happened
  3. Background
  4. The numbers
  5. What it means
  6. What we don't know yet
  7. What to watch next
  8. Sources

AMD reported third-quarter 2026 revenue of $13.377 billion on 3 November, up 45% from a year earlier, and told investors to expect about $17 billion in the fourth quarter, well above what analysts had pencilled in. The headline is the Data Center segment: $8.5 billion, up 97%, now close to two thirds of the company. For developers and anyone watching the AI hardware race, the quarter matters because it is the first one in which AMD says its rack-scale Helios systems, built around Instinct MI450 GPUs, are running real workloads at named customers.

Key points

  • Revenue was $13.377 billion (up 45% year over year, up 16% from the second quarter), against a consensus of about $13.09 billion cited by CNBC.
  • Data Center revenue was a record $8.5 billion, up 97%, "driven by strong demand for AMD EPYC processors and the launch of AMD Instinct MI450 Series GPUs", according to AMD.
  • Non-GAAP earnings per share were $2.09 (GAAP: $1.46); CNBC puts the analyst expectation at $1.95.
  • The fourth-quarter outlook is about $17 billion, plus or minus 3%, with a non-GAAP gross margin of about 55%.
  • Gaming revenue fell 40% to $773 million, and the combined Client and Gaming segment fell 8% to $3.7 billion.

What happened

AMD published its results after the US market closed on 3 November. Third-quarter revenue of $13,377 million compared with $9,246 million in the third quarter of 2025 and $11,536 million in the second quarter of 2026. Gross profit was $7,210 million on a GAAP basis, a 54% margin (52% a year ago). Operating income was $2,832 million, more than double the $1,270 million of a year earlier, and net income reached $2,428 million, up 95%.

On the non-GAAP basis that most analysts track, which strips out stock-based compensation and the amortisation of acquisition-related intangibles, gross margin was 56%, operating income was $3,923 million (a 29% margin) and net income was $3,471 million. Non-GAAP earnings per share grew 74% to $2.09, which CFO Jean Hu said shows "the significant operating leverage" in the business as the data center side scales.

The segment split tells the story better than the totals:

  • Data Center: $8.5 billion, up 97% year over year. CNBC calculates that this is 64% of total revenue, up from 47% a year ago, which fits AMD's own figures.
  • Client and Gaming: $3.7 billion, down 8%. Inside it, the Client business (Ryzen mobile processors) grew 7% to $2.9 billion, while Gaming fell 40% to $773 million on lower semi-custom and graphics revenue.
  • Embedded: $1.1 billion, up 29%, led by aerospace and defense, automotive, communications and industrial customers.

CEO Lisa Su called it "a record quarter with revenue exceeding the high end of our guidance". She said AMD expects growth to accelerate further in the fourth quarter "as we scale Helios and 'Venice' production", Venice being AMD's codename for its upcoming EPYC processors. According to CNBC, she also said she expects server revenue to double year over year in 2027, and that data center AI growth is now accelerating faster than AMD projected 90 days ago.

Background

For most of the generative AI boom, AMD was the credible number two. Its Instinct accelerators sold in the billions but trailed Nvidia by a wide margin, and software (the ROCm stack that rivals CUDA) was the usual complaint from developers. The company's answer was to stop selling only chips and start selling whole racks: the Helios design pairs MI450-series GPUs with sixth-generation EPYC CPUs and networking in a single rack-scale system, the same level at which Nvidia sells its NVL72 systems. Our look at Nvidia's Vera Rubin NVL72 spec page shows the kind of rack AMD is now competing against.

The context for this quarter is a market that has been nervous. CNBC notes AMD's shares had fallen three days in a row before Tuesday's rally, during an industry-wide sell-off driven by fears of AI overspending. That nervousness is why customer evidence matters more than usual: investors want to see that the spending turns into deliveries. For a contrasting quarter, with a fourth-quarter guide below the third quarter's revenue, see our Intel Q3 article.

The numbers

The chart below shows total revenue over the last year's reported quarters and the fourth-quarter guide. The Q4 bar is the midpoint of AMD's own range, not a result.

AMD quarterly revenue (Q4'26 is the guidance midpoint)US$ billion
AMD quarterly revenue (Q4'26 is the guidance midpoint) (US$ billion)
US$ billion
Q3'259.25
Q2'2611.54
Q3'2613.38
Q4'26 guide17.0

Source: AMD Q3 2026 results release, 3 November 2026

The guide implies growth of roughly 66% year over year and 27% from the third quarter, according to AMD. With the plus or minus 3% band, the range runs from about $16.5 billion to $17.5 billion. CNBC cites an average analyst estimate of $16.2 billion from LSEG, so the midpoint sits about $0.8 billion above consensus.

The segment mix is where the growth sits:

AMD Q3 2026 revenue by businessUS$ billion
AMD Q3 2026 revenue by business (US$ billion)
US$ billion
Data Center8.5
Client2.9
Embedded1.1
Gaming0.77

Source: AMD Q3 2026 results release

Gross margin is the one number that moves the wrong way. AMD guides fourth-quarter non-GAAP gross margin to about 55%, from 56% in the third quarter; CNBC says the average estimate was 55.3%, citing StreetAccount. AMD's release does not explain the one-point step down.

What it means

For AI builders. The release names Anthropic, Meta, Microsoft, OpenAI and Oracle Cloud Infrastructure as running models and workloads on the first Helios racks, and HPE said Vultr will deploy the first Helios by HPE systems. CNBC adds that Meta said its Helios racks are up and running. For developers, a second large-scale source of accelerators is mostly good news: it eases supply pressure and strengthens the case for writing code that is portable across vendors. AMD also says it released ROCm 10, with what it calls AI-native developer experiences, though we have not tested it.

For the local-AI crowd. AMD says its Ryzen AI Halo platform with Ryzen AI Max PRO 400 Series processors is now available, and that HP, Lenovo, Framework and GMKtec have expanded availability. It also showed a Threadripper-based workstation it says can run trillion-parameter models locally. Those are the company's claims; we have not seen independent benchmarks.

For the industry. Data Center nearly doubling while Gaming drops 40% shows how far the company has tilted toward AI infrastructure. It also shows concentration risk: a business that depends on a handful of hyperscale and AI-lab customers is exposed to the same spending-fear cycle that rattled markets this week.

Deals. AMD says it agreed to acquire World Labs and Taalas, the latter adding specialised inference silicon. CNBC puts the World Labs price at $8.2 billion in an all-stock deal, announced in September, and reports that AMD has committed up to $5 billion to invest in Anthropic and $1 billion in Meta's equity offering. We have taken those three figures from CNBC only; AMD's release does not state them.

What we don't know yet

  • How much of the $8.5 billion is MI450 and Helios. AMD's release attributes the growth to EPYC and the MI450 launch but does not split GPU revenue from CPU revenue in the text we reviewed.
  • Customer volumes. "Running workloads" is not the same as large orders, and AMD has not published rack counts or per-customer revenue.
  • The margin path. The 55% guide is a single quarter; how rack-scale mix affects margins through 2027 is not stated.
  • Independent performance data. Comparisons with Nvidia's Vera Rubin systems rest on vendor figures so far.
  • Third-party figures. Some analyst sites quote segment margins and cash-flow numbers that do not appear in AMD's release; we left them out.

What to watch next

  • AMD's fourth-quarter report, which will show whether the roughly $17 billion guide holds and whether the gross margin lands near 55%.
  • Evidence of Helios volume: customer announcements, cloud instance availability and rack shipments.
  • The closing of the World Labs and Taalas acquisitions and what they mean for AMD's software and inference roadmap.
  • Nvidia's and the hyperscalers' own spending plans, which set the ceiling for everyone selling into this market.
  • Independent benchmarks of MI450 and Helios against Nvidia's current systems.

Sources

Written by

audiotale.net team

The official account of the people who run audiotale.net: announcements, conversation starters and the newsroom's drafts. Posts written with AI say so. Every other account here is a real person.