
Key Points
- Strong earnings raise expectations as AMD’s 3Q26 revenue grew 45% YoY, while management raised its 2027 revenue growth outlook to above 80%.
- AMD’s competitive position strengthens as EPYC share gains and initial Helios deployments reinforce its role as a credible Nvidia alternative, although further upside requires delivery above plan.
- Supply constraints limit upside surprises as AMD’s USD48.8Bn operating commitments support its 2027 ramp, but tight HBM4, packaging and rack-integration capacity constrain incremental shipments.
- We maintain NEUTRAL with a higher USD573 target price, as the 12.7% upside does not sufficiently compensate for execution risks, customer concentration and potential warrant dilution.
AMD’s 3Q26 results strengthened our fundamental view, but not enough for us to upgrade the stock. Revenue rose 45% year on year to USD13.38Bn, Data Centre revenue increased 97% to USD8.55Bn and non-GAAP EPS rose 74% to USD2.09, while Helios moved from roadmap to recognised revenue and live customer workloads. Management also materially raised its 2027 outlook. The issue is that the expectations bar has risen with the fundamentals. AMD now needs to deliver above an already aggressive plan, yet the incremental HBM, packaging and system capacity required to create that surprise remains constrained. We maintain NEUTRAL and raise our target price to USD573 from USD509.
Figure 1: AMD’s share price

Earnings summary
AMD delivered a modest headline beat with better earnings quality. Revenue increased 45% year on year and 16% sequentially to USD13.38Bn, 2.3% above consensus, while non-GAAP EPS rose 74% year on year to USD2.09, 6.1% ahead of consensus. Data Centre remained the key driver, with revenue up 97% year on year to USD8.55Bn and segment operating income up 168% to USD2.88Bn. Group non-GAAP operating margin expanded around 510bps year on year to 29.3%.
Free cash flow fell -64% year on year to USD554Mn, below consensus of USD1.97Bn, as AMD built USD3.4Bn of working capital. Inventory rose USD3.0Bn, mostly raw materials and work in process, to secure 2027 supply. On our estimates, the build is sized for management’s raised plan rather than above it, covering the 4Q26 ramp and the guided 1Q27 step-up, while USD22.2Bn of commitments due in 2027 equals about half of our FY2027 cost of goods. This supports delivery of the plan, but not upside beyond it, as Helios racks still need more memory and third-party components. Cash conversion should stay weak, and 4Q26 free cash flow could turn negative.
Table 1: AMD’s 3Q26 earnings
|
Metric |
3Q26 |
2Q26 |
3Q25 |
QoQ |
YoY |
Consensus/Guide |
Beat/Miss |
|
Revenue (USD mn) |
13,377 |
11,536 |
9,246 |
16.00% |
44.70% |
Consensus USD13,080mn; guide USD12.7–13.3bn |
Beat (+2.3% vs consensus; +2.9% vs guide midpoint) |
|
Non-GAAP gross margin |
56.00% |
56.00% |
54.00% |
Flat |
+200bps |
Consensus 56.1%; guide ~56% |
In line (-10bps vs consensus) |
|
Non-GAAP operating margin |
29.30% |
26.80% |
24.20% |
+250bps |
+510bps |
Consensus ~28.0% |
Beat (+130bps) |
|
Non-GAAP EPS (USD) |
2.09 |
1.66 |
1.2 |
25.90% |
74.20% |
Consensus USD1.97 |
Beat (+6.1%) |
|
Data Centre revenue (USD mn) |
8,546 |
6,718 |
4,341 |
27.20% |
96.90% |
Consensus USD8,200mn |
Beat (+4.2%) |
|
Data Centre operating income (USD mn) |
2,882 |
2,103 |
1,074 |
37.00% |
168.30% |
No guide or consensus |
Margin 33.7% (+900bps YoY) |
|
Client and Gaming revenue (USD mn) |
3,722 |
3,841 |
4,048 |
-3.10% |
-8.10% |
Consensus USD3,715mn (Client USD3,060mn + Gaming USD655mn) |
In line (+0.2%): Client -3.6%, Gaming +18.0% |
|
Embedded revenue (USD mn) |
1,109 |
977 |
857 |
13.50% |
29.40% |
Consensus USD1,090mn |
Beat (+1.7%) |
|
Operating cash flow (USD mn) |
1,202 |
2,366 |
1,788 |
-49.20% |
-32.80% |
No guide or consensus |
– |
|
Capex (USD mn) |
648 |
808 |
258 |
-19.80% |
151.20% |
Consensus USD647mn |
In line (+0.2%) |
|
Free cash flow (USD mn) |
554 |
1,558 |
1,530 |
-64.40% |
-63.80% |
Consensus USD1,970mn |
Miss (-71.9%) |
|
Source: Bloomberg Finance L.P., iFAST compilations. Data as of 4 November 2026. |
|||||||
Figure 2: AMD’s inventory build is wafers in the line, not unsold chips

Source: AMD
Management guidance
Guidance was more important than the quarterly beat. AMD guided 4Q26 revenue to USD17.0Bn ±3%, around 4% above consensus and up 66% year on year at the midpoint. Data Centre revenue is expected to rise around 50% sequentially to approximately USD12.8Bn, while non-GAAP gross margin is guided to around 55%. For 2027, management expects group revenue growth above 80%, server CPU revenue to roughly double and Data Centre AI revenue to significantly more than double.
Table 2: Management guidance
|
Metric |
Management guide |
Consensus |
Implication |
|
4Q26 revenue |
USD17.0Bn ±3% |
USD16.3Bn |
Beat Street by 4% |
|
4Q26 non-GAAP gross margin |
55.0% |
55.5% |
50bps below consensus |
|
4Q26 non-GAAP EPS |
USD2.92 |
USD2.68 |
9% above consensus |
|
4Q26 Data Centre revenue |
USD12.8Bn |
- |
50% QoQ growth |
|
2027 revenue growth |
>80% (Aug: "substantially above" the >35% long-term target) |
77% |
Above consensus; implies FY2027 revenue above USD93.9Bn |
|
2027 server CPU revenue growth |
~100% (Aug: >70%) |
- |
Raised from >70% in August; mostly unit-driven, with some ASP uplift from higher core counts |
|
2027 Data Centre AI revenue |
Significantly more than double; faster than projected 90 days ago |
- |
Raised |
|
2027 gross margin |
Not guided; 4Q26 (~55%) as benchmark |
- |
Depends on GPU vs server CPU mix |
|
2027 PC market units |
Down double digits YoY |
- |
Headwind for Client |
|
Source: Bloomberg Finance L.P., iFAST compilations. Data as of 4 November 2026. |
|||
AMD rose as much as 7% overnight to about USD519, but most of the gain faded as US futures fell on the midterm election results, and the shares closed only 1.0% higher at USD488.78. We read this as the stronger outlook being enough to support existing earnings expectations, but not yet enough to drive another substantial rerating.
AMD then fell -17.3% on 5 November, roughly twice Nvidia’s -9.3%, on fears that newly elected Democrats would slow data-centre construction. As argued in “Post-Midterm AI Sell-off: Data Centre Curbs Are a Headwind, Not the End of the Trade”, this was a positioning shock rather than a demand shock, with state permitting more likely to slow the build-out than stop it. AMD fell harder because Helios depends on new sites coming online. The shares rebounded +11.0% and +13.3% over the next two sessions to USD508.58, helped by an Oppenheimer upgrade and Google’s EPYC win. Demand confidence is intact, but valuation is now highly sensitive to deployment delays.
Helios has passed the first delivery test
In our August update, we argued that “AMD remains fundamentally well positioned, but the investment case has shifted from whether it can deliver strong growth to whether it can materially exceed an already elevated market bar.” The latest results reinforce this view, with AMD making tangible progress in both server CPUs and AI accelerators.
EPYC remains AMD’s most established growth engine. Cloud and enterprise server sales each more than doubled in 3Q26, while management raised its 2027 server revenue outlook from more than 70% growth to roughly 100%. The rollout of Venice further supports AMD’s ability to capture share from Intel, providing a more established earnings foundation alongside its AI accelerator ambitions.
Helios has also cleared an important milestone. MI455X and Helios generated their first revenue in 3Q26, with Anthropic, Meta, Microsoft, OpenAI and Oracle running workloads on initial systems. This reduces the product-readiness concerns raised in August and strengthens AMD’s credibility as an alternative to Nvidia. Customers also have strategic incentives to diversify their accelerator suppliers, giving AMD room to expand without displacing Nvidia's market leadership.
However, stronger product credibility does not automatically translate into further share-price upside. EPYC share gains, Helios adoption and AMD’s growing AI pipeline are increasingly reflected in expectations. The next test is no longer whether AMD can deliver competitive products, but whether it can scale shipments fast enough to exceed management’s already aggressive 2027 outlook. That remains unproven.
The next rerating needs delivery above plan
The midpoint of 4Q26 revenue guidance takes our FY2026 base to approximately USD52.2Bn. Management’s >80% growth framework consequently implies more than USD93.9Bn in 2027, close to our revised USD94.5Bn forecast. Management tend to guide conservatively, having beaten its revenue guidance midpoint in each of the past 11 quarters, but the size of the beat has narrowed from +7.0% in 4Q25 to about +3% in the past two quarters as supply tightened, so modest beats are largely expected. Delivering that result would be exceptional growth, but not a major surprise. The first test is 4Q26 Data Centre revenue above the implied USD12.8Bn while gross margin remains near 55%, followed by repeatable gigawatt-scale shipments and customer acceptance through 2027.
Table 3: Helios delivery milestones and positive-surprise tests
|
Window |
What is currently expected |
Evidence needed for upside |
|
3Q26 |
Initial MI455X / Helios revenue |
Achieved; five lead customer workloads |
|
4Q26 |
Material volume ramp |
Data Centre >USD12.8Bn at ~55% GM |
|
1H27 |
OpenAI, Meta and Anthropic deployments |
Faster gigawatt commissioning and acceptance |
|
2027–28 |
Repeat orders and MI500 follow-on |
Volume commitments beyond current supply cycle |
|
Source: Bloomberg Finance L.P., iFAST compilations. Data as of 4 November 2026. |
||
Supply supports the plan, but caps how far AMD can surprise
Management’s 2027 growth framework appears supported by planned component supply, reducing the risk of a major delivery shortfall. However, exceeding that framework requires incremental HBM, advanced-packaging and substrate capacity beyond current allocations. AMD’s operating commitments have quadrupled since end-2025, strengthening confidence in its planned ramp, but these obligations do not establish how much incremental Helios capacity it can secure.
Figure 3: AMD’s purchase commitments have quadrupled since end-2025 as it secures supply for the 2027 ramp

HBM4 is a critical constraint because more GPU wafers will not become finished MI455X accelerators without matching memory. Advanced packaging capacity is also contested, while substrates can hold back both Instinct and EPYC. Rack-scale delivery introduces another layer. A Helios system can only be shipped and accepted once accelerators, CPUs, networking, power and cooling are available together. Customer-site construction and power availability can delay deployment even after the hardware is assembled.
Easing memory shortages would make delivery surprises more possible, but the evidence points to continued tightness. Samsung has sold out its 2027 HBM, SK hynix has contracted more than half of its 2027–28 capacity with further volumes under negotiation, and Micron expects 2027 and 2028 to be tighter than 2026. New capacity is therefore being allocated in advance, with AMD competing against Nvidia and custom-chip developers. AMD has shown it can secure meaningful supply in this market, through its higher commitments and its HBM4 agreement with Samsung, but neither company has disclosed how much capacity AMD holds beyond its current plan, and only about USD5Bn of AMD’s commitments fall due in 2028. The risk of AMD missing its 2027 ramp has fallen, but the evidence for deliveries above guidance is not yet there.
Table 4: Where AMD's delivery could bind
|
Constraint |
Current assessment |
Implication for production or revenue |
|
Leading-edge wafers |
Planned for current outlook |
Incremental wafer allocation is potential upside |
|
HBM4 |
Important incremental-supply risk |
Extra GPU output needs matching memory |
|
Advanced packaging |
Industry capacity remains tight |
Completed chips may lag wafer availability |
|
Substrates |
Industry-wide pressure |
Potential bottleneck for EPYC and Instinct |
|
Rack integration |
Early commercial ramp |
One missing component delays full-system delivery |
|
Customer site readiness |
Deployment-linked uncertainty |
Power and commissioning can defer acceptance |
|
Source: AMD, iFAST compilations. Data as of 4 November 2026. |
||
AMD reported USD48.8Bn of unconditional operating commitments in 3Q26, with USD33.4Bn due by the end of 2027. These commitments extend beyond semiconductor supply purchases and should not be read as guaranteed chip-production capacity. Inventory increased 35% sequentially to USD11.46Bn as AMD prepared for the ramp; if deployments slip, that working-capital build becomes a warning sign rather than evidence of available component capacity.
Demand visibility comes with financing exposure
The pipeline is also concentrated. In our August tally, OpenAI and Meta accounted for approximately 12GW of 14.6GW in announced commitments, deployed over multiple years through 2030, with roughly 3GW expected across customers in 2027. Both are developing their own accelerators, leaving open how much of today’s AMD share will persist once Nvidia supply improves and custom silicon expands. The customer arrangements support genuine demand, but they do not guarantee lasting merchant-GPU market share.
AMD is also supporting some infrastructure demand financially. Its 3Q26 filing reports up to USD8.5Bn in guarantees of partners’ lease and debt obligations, up from approximately USD4.1Bn in June, alongside USD13.3Bn of future data-centre leases intended for assignment or sublease, USD6.2Bn of cloud-compute commitments and up to USD5.1Bn of equity investment commitments. These are distinct maximum exposures, not cash already spent or amounts that should be summed as debt. Management reports no partner defaults. On the call, questions about financing were answered mainly with confidence in deployment-site visibility, leaving commercial risk allocation an important monitoring point. Our take is that this reduces quality of the revenue, limiting us from further revising our fair P/E upwards.
OpenAI and Meta also each hold warrants to acquire up to 160Mn AMD shares at USD0.01, subject to deployment, share-price and other performance conditions. While these arrangements support AMD's long-term demand visibility, they also mean successful deployments could trigger substantial share dilution, reducing the earnings upside attributable to existing shareholders.
Risk
The main downside is a delivery miss against the higher 2027 bar, which could lower earnings and compress the multiple simultaneously. Helios gross-margin pressure, weaker cash conversion, export restrictions or a hyperscaler capex pause would compound this. Intel’s 18A/Clearwater Forest roadmap and Nvidia’s integrated systems also threaten share gains. We would become more constructive on Data Centre sales above USD12.8Bn in 4Q26 at roughly 55% gross margin, incremental HBM and packaging supply, stronger free cash flow and repeat MI500 commitments beyond the initial customer cycle.
Valuation
We raise FY2026E revenue to USD52.2Bn, up 51% year on year, and FY2027E to USD94.5Bn, up 81%. FY2028E revenue growth moderates to 24.4% as the Helios base effect normalises and competition from Nvidia and custom accelerators intensifies.
Our warrant-adjusted EPS of USD15.91 for FY2027 and USD19.11 for FY2028 includes illustrative additional diluted shares of 67Mn and 133Mn respectively. We retain a 30x FY2028E fair P/E, near the five-year average of approximately 31x. AMD still commands a premium to larger AI-chip peers, supported by faster forecast earnings growth from a smaller base, but we assume no further scarcity-driven rerating. The resulting USD573 target implies 12.7% upside from USD508.58.
We retain NEUTRAL because this return depends on FY2028 earnings and successful execution of the full 2027–28 ramp. At the same 30x multiple, FY2027E EPS alone implies approximately USD465 per share, illustrating how much of the target is contingent on further delivery. In the absence of clearer evidence of incremental capacity, system economics and dilution, the risk-adjusted upside is insufficient for an upgrade.
Table 5: Valuation
|
|
FY2025 |
FY2026E |
FY2027E |
FY2028E |
|
Revenue (USD Mn) |
34,639 |
52,166 |
94,478 |
117,519 |
|
Revenue Growth (%) |
|
50.6% |
81.1% |
24.4% |
|
P/E (x) |
121.9x |
63.2x |
32.0x |
26.6x |
|
Warrant-adjusted non-GAAP EPS |
4.17 |
8.05 |
15.91 |
19.11 |
|
EPS growth (%) |
|
93.0% |
97.5% |
20.1% |
|
Fair P/E (x) |
|
|
|
30x |
|
Current price (USD) |
|
|
|
508.58 |
|
Target price (USD) |
|
|
|
573 |
|
Upside potential (%) |
|
|
|
12.7% |
|
Source: Bloomberg Finance L.P., iFAST compilations. Data as of 10 November 2026. |
||||
Key takeaway
AMD’s fundamentals have strengthened over the past three months, supported by continued EPYC share gains, initial Helios deployments and a substantially stronger 2027 outlook. However, much of this progress is already reflected in market expectations, leaving AMD with a high bar for further earnings surprises. While the company is increasingly established as a credible alternative to Nvidia, incremental supply constraints limit its ability to accelerate deliveries beyond management’s plan. We therefore raise our target price to USD573 to reflect stronger earnings forecasts but maintain NEUTRAL as the potential upside relies on successful execution of an aggressive growth trajectory.

