AMD vs Amazon: Which AI stock has the stronger 2027 payoff?

Oracle AI World is opening in Las Vegas with a question that matters more to investors than the size of the keynote schedule: who captures the cash when enterprise AI moves from demonstrations into production? AMD sells the compute that makes the workload possible. Amazon sells the cloud capacity, custom chips and software layer that turns that compute into a recurring service. The better stock depends on which part of that chain converts demand into durable profit.

Hero image: Blue-lit server racks. Photo via Unsplash.

The timing gives the comparison a live news hook. Oracle's official AI World agenda places the main event from October 25 through October 28, with an October 26 keynote and partner sessions built around enterprise AI, infrastructure and measurable outcomes. AMD is a visible infrastructure partner at the event, while Amazon's AWS competes for the same enterprise workloads through cloud services and its own silicon. This is a dated comparison tied to an event now under way, not a claim that either company has issued new guidance today.

The decision in one sentence

AMD offers the cleaner operating leverage to a second wave of accelerator demand, but Amazon offers the stronger cash-conversion cushion because AWS already turns AI usage into a large, profitable recurring business. For a 2027 investor, AMD is the higher-upside execution bet; Amazon is the more forgiving way to own the AI build-out if hardware cycles or customer concentration disappoint.

That conclusion comes from a calculation built from the companies' latest comparable disclosures. AMD reported $6.7 billion of second-quarter 2026 Data Center revenue, up 107% year over year, in its August 4 results release. Amazon reported $42.232 billion of AWS revenue in the second quarter and a $169 billion annualized run rate in its July 30 results. Annualizing AMD's quarter produces $26.8 billion of Data Center revenue. Dividing that by AMD's roughly $857.4 billion market value, compared with Amazon's $169 billion AWS run rate divided by roughly $2.704 trillion of market value, gives an AI-revenue-to-market-cap ratio of about 3.1% for AMD versus 6.3% for Amazon. Amazon's ratio is about twice AMD's even though AMD's segment is growing much faster.

That is the standout test in this comparison: growth is not the same as monetization density. AMD needs its roadmap and customer commitments to become many quarters of recognized revenue. Amazon already has the billing relationship, software attach and operating cash flow.

What Oracle AI World changes

An event is useful only if it changes the evidence investors will watch next. Oracle's 2026 program is unusually focused on the gap between an AI demo and a production workflow. The keynote page says the company will focus on putting AI into trusted data, governed systems and the processes that run a business. The partner agenda also includes sessions on multicloud infrastructure, AI marketplaces and turning demand into revenue.

For AMD, that is a demand-qualification test. The company has announced a broad rack-scale platform rather than a single accelerator. Its Advancing AI 2026 release describes Helios racks built from 72 MI455X GPUs, 18 sixth-generation EPYC CPUs, Pensando networking and ROCm software. AMD says Helios can deliver up to 30% more inference tokens per dollar than the leading competing solution, while also naming OpenAI, Anthropic, Meta, Microsoft and Oracle among collaborators. Those are company claims and road-map commitments; they become investment evidence only when customers deploy systems and AMD reports the revenue, margins and receivables that follow.

For Amazon, AI World is a distribution test. AWS can sell an enterprise customer a managed service, a database connection, model access, security controls and custom silicon in one account. Amazon's second-quarter filing says AWS revenue rose 37% year over year and operating income rose 64% to $16.621 billion. The same filing also says AWS and Anthropic expanded their strategic collaboration and multi-year commitment by more than $100 billion over ten years, including contractual obligations related to AWS chips. That is a significant demand signal, but it also shows why the stock must be judged on capital intensity: Amazon is spending to build the capacity before the revenue arrives.

The scorecard

Revenue engine. AMD is the specialist. Data Center represented 58% of its second-quarter company revenue, and the segment more than doubled year over year. Amazon is the diversified platform. AWS is 21% of second-quarter sales, but its operating profit is large enough to fund investment across the rest of the company. AMD wins the growth-rate test; Amazon wins the recurring-revenue test.

Margin quality. AMD reported a 54% GAAP gross margin and $1.99 billion of operating income in the quarter. Amazon's AWS produced $16.621 billion of operating income on $42.232 billion of sales, a 39.3% segment operating margin. The measures are not identical because AMD is a hardware segment and AWS is a service segment, but the gap explains the stock risk: AMD can grow faster while Amazon can absorb more spending without breaking consolidated earnings.

Capital risk. AMD's model is fabless, so it avoids owning the full manufacturing footprint, but it must secure advanced packaging, memory and foundry capacity. Amazon owns the demand relationship and much of the data-center build-out. Its 10-Q makes the trade-off explicit: the company is signing long commitments and expanding infrastructure while AWS demand accelerates. Investors get more visibility into recurring usage, but they also take depreciation, power and financing risk.

Customer power. AMD's named partners are impressive, yet the company remains dependent on a small group of hyperscalers and AI labs for the largest deployments. Amazon has the opposite position: it is a supplier to many customers, but a few frontier-model customers can negotiate hard on price and capacity. AMD's concentration risk is higher at the chip level; Amazon's bargaining risk is higher at the cloud level.

Valuation. A simple trailing P/E comparison from the latest market snapshot is not a complete valuation model, but it highlights the hurdle. AMD traded near $516.79 with a market value near $857.4 billion and a trailing P/E around 132.9. Amazon traded near $248 with a market value near $2.704 trillion and a trailing P/E around 20.0. These prices were measured at the October 23, 2026 close in the market feed, so they are a dated reference rather than a live quote. AMD's multiple assumes a steep earnings ramp; Amazon's multiple assumes that AWS growth and retail cash flow can fund the AI build-out.

A reproducible 2027 forecast

The forecast is a scenario test, not a price target. I hold each company's market value constant to isolate the operating hurdle.

For AMD, the annualized Data Center base is $26.8 billion. In a bear case, 2027 Data Center revenue reaches $32 billion as customer qualifications slip and accelerator pricing compresses. In a base case, it reaches $45 billion as Helios and MI-series deployments move from pilot to production. In a bull case, it reaches $60 billion as the named hyperscaler relationships scale. At the current market value, those scenarios imply Data Center revenue-to-market-cap ratios of 3.7%, 5.2% and 7.0%. The base case requires roughly 68% growth from the annualized second-quarter run rate. That is ambitious, but it is consistent with a company that has already delivered 107% year-over-year segment growth; the risk is that the growth rate decelerates before the valuation does.

For Amazon, I start with the $169 billion AWS annualized run rate. A bear case puts 2027 AWS revenue at $185 billion, a base case at $210 billion and a bull case at $240 billion. At the current market value, those figures imply AWS revenue-to-market-cap ratios of 6.8%, 7.8% and 8.9%. The base case is a 24% increase from the latest run rate, slower than AWS's 37% second-quarter growth but still strong for a business already operating at scale. Amazon does not need to prove that AI demand exists; it needs to prove that incremental capacity earns an acceptable return after depreciation, power and chip commitments.

The most useful signal from Oracle AI World is therefore not a new product name. It is evidence that customers are moving from isolated pilots to governed, multi-step workloads. If AMD leaves the event with named production deployments and subsequent backlog conversion, its higher multiple can be defended. If Amazon leaves with more enterprise workloads attached to AWS services and custom chips, its lower multiple can rerate without heroic assumptions.

Which stock wins, and what would change the call?

AMD wins for an investor who accepts execution risk and wants direct exposure to the accelerator shortage. The upside is not merely selling more chips. It is selling a rack-scale system, networking and software stack that captures more of the customer's budget. AMD's own materials describe that strategy, and the 2027 MI500 roadmap gives the market a visible next step. The warning sign would be a widening gap between announced gigawatts and recognized Data Center revenue, or a margin decline that shows AMD is buying growth with discounts.

Amazon wins for an investor who values cash-flow durability. AWS already has a $169 billion run rate and $16.6 billion of quarterly operating income. Its custom-chip commitments can lower cost and improve availability, while Bedrock and adjacent services create software revenue around the hardware. The warning sign would be AWS growth falling sharply while capital expenditure stays elevated, forcing investors to wait longer for returns. Amazon's scale makes the disappointment less dramatic than AMD's, but it also makes explosive upside harder.

My forecast verdict is split: AMD has the larger percentage upside if the 2027 deployment curve is real; Amazon has the better risk-adjusted setup if the AI cycle becomes more capital-intensive and less profitable than the headlines suggest. The decision should change after three checks: AMD's next report should show whether Data Center growth is converting into cash and stable margins; Amazon's October 29 report should show whether AWS growth is holding while capex rises; and Oracle AI World's customer sessions should reveal whether enterprises are buying production capacity or still shopping for pilots.

That is why this is a real versus decision rather than a generic “best AI stocks” list. AMD is a bet on the next unit of compute. Amazon is a bet on owning the meter that charges for every unit after it is deployed.