AMD’s AI Boom: Can Chip Demand Push the Stock to New Highs?

AMD’s AI opportunity is becoming more tangible, but its stock has already moved a long way ahead. Lisa Su’s latest comments describe demand strong enough to require years of supply-chain planning. My view is that AMD has a credible route to much larger earnings, while further sustained stock gains depend on delivering enough profit per share to justify the expectations already in the price.

Let's break down what Lisa Su has just said, how AI demand reaches AMD’s business and what its valuation suggests about the path to new highs.

What has Lisa Su just said about AMD’s AI chip demand?

Speaking to reporters in Taipei on October 6, AMD’s chair and CEO said the company would substantially increase supply in 2027. Reuters reported that AMD was planning capacity three to five years ahead because demand remained “very, very high.”

Her remarks were about the work needed to expand CPU and GPU production. They were not a new numerical revenue forecast or a guarantee that every customer request would become a shipment.

The message is encouraging for demand visibility. It also explains why investors should track available supply and delivery schedules alongside orders.

The Korea Herald also reported that Su said next-generation MI455 accelerators and Helios systems had begun shipping with HBM4 memory from Samsung and SK hynix. That indicates progress on the product ramp, but initial shipments do not establish how quickly full-scale customer deployments will grow.

High-bandwidth memory (HBM) supplies data rapidly to AI processors. Samsung’s earlier official announcement described collaboration on HBM4 for AMD’s MI455X accelerator and memory for EPYC processors. Su’s current supply-chain tour builds on that existing relationship.

Focus Taiwan/CNA reported another useful update from Taipei: Su said Helios shipments had begun in the third quarter as planned. She also said AMD’s previously announced $10 billion Taiwan supply-chain investment was progressing and that investment would increase, without specifying a revised amount.

Samsung’s announcement also mentioned discussions about possible foundry cooperation. Discussions should not be presented as a confirmed new manufacturing contract.

The distinction is useful: demand can give a supplier confidence to expand, but capacity still has to be built and qualified before it produces revenue.

Why does AI demand help both AMD GPUs and CPUs?

AMD has more than one route into AI infrastructure. Its Instinct GPUs handle demanding AI calculations, while EPYC CPUs support the broader computing work around them. Its ROCm software helps applications use AMD hardware.

AMD’s Helios reference design combines accelerators, CPUs and networking in a coordinated rack. It is a blueprint for partners to build systems rather than a complete AMD-branded rack sold directly through that product page.

That matters financially because AMD can participate in several parts of a customer’s computing budget. It also means a headline about AI GPU demand does not capture the whole opportunity.

AI agents could broaden that opportunity further. An agent may perform several steps to complete a task, creating work for processors beyond the calculations performed by an AI model. How much extra demand this creates depends on actual usage and the way customers design their systems.

Large announced customer agreements provide more concrete evidence than a market-size projection alone.

These are multiyear deployment agreements. A gigawatt describes infrastructure scale rather than a dollar amount of AMD revenue and the announced starting dates do not establish that the full agreements have already been delivered.

Sources: AMD’s OpenAI partnership announcement, 6 October 2025; Meta partnership announcement, 24 February 2026; Anthropic partnership announcement, 22 July 2026. 

The commercial significance is that major customers are designing AMD into their infrastructure plans. Engineering collaboration can make a platform more useful for specific workloads, which can help turn an initial deployment into repeat demand.

The financial test remains delivery at attractive margins. Announced capacity should never be converted mechanically into booked revenue without knowing the equipment mix, pricing and shipment schedule.

Do AMD’s latest earnings support the AI growth story?

The latest published quarter is fiscal Q2 2026, which ended on June 27. AMD’s results show that the business expansion is already visible in revenue rather than existing only in future announcements.

Data Center is now the largest part of AMD’s revenue. However, the segment includes server CPUs and other products as well as AI accelerators, so its total should not be described as AI GPU sales.

Sources: AMD’s Q2 2026 results release and Form 10-Q; Q2 earnings slides filed with the SEC. 

The profit comparison also needs context. The previous year’s quarter included an $800 million inventory and related charge associated with US export restrictions on MI308 products. Some of the improvement in the year-over-year margin comparison reflects the absence of that charge.

GAAP figures follow standard accounting rules. Adjusted figures exclude specified items such as stock-based compensation and acquired-intangible amortisation. The adjusted result helps assess operating trends, but those exclusions should still be considered when judging shareholder economics.

Free cash flow adds another useful check because accounting profit and cash generation can move differently. AMD’s reported figure deducts purchases of property and equipment from operating cash flow from continuing operations.

Management’s next-quarter outlook provides the near-term benchmark.

This is guidance rather than a reported result. It sets a clear test for whether the second-half expansion is reaching the income statement.

Source: AMD’s Q2 2026 earnings announcement, “Current Outlook.”

My assessment is that AMD’s operating momentum is substantial. The more demanding question is whether that momentum can exceed expectations that have risen alongside the stock.

How does AMD compare with Nvidia in AI chips?

AMD does not have to overtake Nvidia to build a much larger business. It needs to secure profitable deployments and repeat customer spending in a market that is expanding.

The latest quarterly results nevertheless show the scale of the competitive challenge.

These periods do not end on the same date and the companies’ Data Center segments contain different product mixes. The comparison illustrates business scale and profitability rather than a precise measure of AI accelerator market share.

Sources: AMD’s Q2 2026 financial statements; Nvidia’s Q2 fiscal 2027 results announced on 26 August 2026.

Nvidia’s scale gives customers a substantial established platform to consider. AMD’s opportunity is to demonstrate that its hardware and software work well for the workloads customers actually run.

The relevant customer calculation is broader than chip price. It includes how much useful work a system completes, the power it consumes and the effort needed to deploy and maintain it. Hardware specifications alone cannot establish the winner.

AMD’s open reference-design approach is a plausible competitive advantage for customers seeking flexibility. That is an analytical judgement, not proof that AMD will achieve a particular market share or match Nvidia’s margins.

A larger market can support growth at both companies. For AMD shareholders, the important outcome is profitable expansion per share.

What does AMD’s stock valuation already assume?

AMD closed at $649.42 on October 6, up 2.80% that day. Investopedia reported that it had reached a fresh intraday record during the session. The question is therefore whether earnings can support further highs after an already strong advance.

The two multiples use the same stock price but different forecast years. The lower 2027 figure depends on a much larger earnings number actually arriving.

A price-to-earnings multiple tells us how much investors are paying for each dollar of annual earnings. Here, even the lower multiple requires the estimated profit per share to more than double between the two forecast years.

These are adjusted earnings estimates rather than GAAP results or AMD guidance. They can change as analysts update shipment, margin and spending assumptions.

Recent target increases show why the market is optimistic, but targets are analyst opinions rather than future prices.

Both targets sit above the closing price, although they imply different amounts of further upside. They do not establish a single dependable outcome.

My view is that AMD’s valuation already recognises considerable success. Strong demand helps the case, but another sustained advance needs earnings delivery or further upward revisions large enough to offset any reduction in the multiple investors are willing to pay.

What would need to happen for AMD stock to reach further highs?

A useful framework separates three steps: demand reaching shipments, shipments producing profit and profit growing per share.

The first step depends on manufacturing and memory capacity. The second depends on pricing, product mix and costs. The third also depends on the number of shares outstanding.

The stock price adds one more variable: the earnings multiple. A company can grow strongly while its shares disappoint if investors become less willing to pay a high multiple.

Consider the following future-state scenarios. They are illustrative assumptions rather than forecasts or price targets. The EPS figures represent possible annual adjusted earnings in a later year and the table gives values in that future year, not discounted values today.

The middle case is the important lesson: substantial earnings can coexist with a share price close to today’s level. The upside case requires both a larger profit base and investors continuing to assign a premium valuation.

AMD has given investors an ambitious long-term reference point. At its November 2025 Financial Analyst Day, the company outlined targets for the following three to five years including revenue growth above 35% annually and adjusted EPS exceeding $20. Those are management targets rather than assured outcomes.

Timing changes the significance of that target. If the illustrative $900 price were reached four years from now, the gain from the current close would be about 8.5% a year before costs and currency effects. It would not be a 38.6% annual return.

A further sensitivity makes the point clearer. At the 2027 consensus EPS estimate, a 35-times multiple would imply roughly $550 rather than the current closing price. That is a conditional calculation showing how valuation compression could outweigh earnings growth.

The strongest evidence for a durable advance would therefore be improving delivery visibility accompanied by higher earnings estimates and healthy margins. More demand headlines alone would provide less information.

What risks could weaken AMD’s AI earnings growth?

Supply limitations can defer revenue even when customers remain interested. Memory and manufacturing partners may also face competing demand from other chipmakers. Securing capacity is therefore both a growth enabler and a financial commitment.

AMD’s customer agreements carry additional economics that deserve attention.

The warrants are conditional rather than shares already issued in full. The World Labs filing says the final number of acquisition shares depends on a pricing formula near closing, so it should not be calculated as a fixed amount from today’s stock price.

Sources: AMD’s OpenAI, Meta and Anthropic partnership announcements; AMD’s World Labs Form 8-K filed on 28 September 2026.

Dilution does not automatically make a transaction unattractive. The question is whether the additional earnings generated more than compensate for the additional shares. That is why total revenue growth and EPS growth should be tracked separately.

The proposed World Labs transaction could strengthen AMD’s understanding of emerging AI workloads. Its financial benefit still requires evidence. Our World Labs acquisition analysis explains that transaction in more detail.

Customer concentration and financing are another consideration. A large infrastructure agreement can depend on a customer’s ability to fund facilities and equipment. AMD’s own investment commitments make cash generation an important part of the assessment.

Export restrictions are a demonstrated risk rather than a hypothetical concern: the MI308 charge shows how policy can change inventory economics. Competition can also affect pricing even if the wider market continues to grow.

For the global supply chain, AMD’s expansion connects US chip design with Taiwanese manufacturing and Korean memory. Su’s remarks make that dependence clear. Geographic disruptions or slow capacity additions can affect when demand becomes revenue.

What should investors watch next in AMD’s AI business?

AMD announced on October 6 that its Q3 results will be released on November 3 after the US market close. That is the next scheduled opportunity to compare management’s guidance with actual performance.

These measures make the CEO’s demand outlook testable. None requires guessing a stock price from a product announcement.

Sources: AMD’s earnings-date announcement, 6 October 2026; Q2 guidance and financial disclosures. 

For Indian investors, AMD offers exposure to an overseas AI infrastructure business rather than a direct proxy for Indian technology services. The outcome in rupees also depends on USD/INR movements: a stronger dollar can add to a dollar-denominated stock return, while a stronger rupee can reduce it.

That currency effect does not improve AMD’s underlying earnings. It is a separate part of an Indian investor’s return and should not obscure the company analysis.

AMD’s latest comments support a positive view of its business opportunity. My assessment of the stock is more demanding: at its current valuation, execution must deliver substantial earnings growth per share. Further highs are plausible if supply expansion, profitable deployments and forecast upgrades arrive together. The next useful evidence will come from shipments and financial results.