• AMZN is the clearest high-conviction way to participate in AI infrastructure: AWS grew 36.7% year over year, and its 2026 capacity is sold out, with much of 2027 also committed.
• Amazon plans about $220 billion in 2026 capital spending, so monitor AWS growth and free cash flow to see whether the investment is earning adequate returns.
• NVDA remains positioned to benefit from strong AI-computing demand, but track competition from Amazon’s Trainium and other custom chips, which could pressure third-party chip demand.
• Treat Anthropic and OpenAI as private-company exposure, not publicly traded stock opportunities; the cited investment figures were not confirmed.

Amazon.com, Inc. (AMZN)

• CEO Andy Jassy said Amazon expects to spend about $220 billion on capital expenditures in 2026, largely to expand data centers and related infrastructure. Even at that level, he said Amazon expects to leave significant AI demand unmet.
• Amazon’s cloud business, AWS, had a $169 billion annual revenue run rate and grew 36.7% year over year in the most recent quarter discussed. Jassy said AWS’s AI business had passed a $25 billion run rate in its first three years.
• Jassy said AWS capacity for 2026 was sold out, 2027 was largely sold out, and a substantial portion of 2028 was already committed. He expects supply constraints in some components to persist for several years.
• Amazon is developing its own chips through Annapurna Labs. Jassy said its Graviton CPUs offer about 40% better price performance than comparable x86 processors, while its Trainium AI accelerators offer about 30%–40% better price performance than comparable processors.
• Beyond AWS, Amazon is investing in Amazon Leo satellite broadband, Zoox autonomous ride-hailing, healthcare, robotics, and AI-driven updates to consumer experiences such as Alexa.
• Jassy described the investments in Anthropic and OpenAI as strategic partnerships, citing customer demand and collaboration on cloud and chip services. The interviewers mentioned estimates of $18 billion for Anthropic and $50 billion for OpenAI, but Jassy did not confirm those amounts.
• Risks mentioned: The build-out can weigh on free cash flow in the near term because data centers require investment well before they generate revenue. Amazon also faces constraints in power, cooling, memory, chips, and data-center capacity, as well as local political resistance to data-center development.

Takeaways

• Amazon offers exposure to several AI-related opportunities—cloud computing, infrastructure, custom chips, and consumer applications—while also having large non-AI businesses.
• The scale of planned investment is a major part of the growth thesis, but it also makes capital spending, free cash flow, and the pace at which new capacity is monetized important factors to track.
• Watch whether AWS growth and AI demand translate into returns on the build-out, and whether Amazon’s chips and other newer businesses gain meaningful customer adoption.

NVIDIA (NVDA)

• Jassy described NVIDIA as a major provider with which Amazon has a deep partnership that he expects to continue.
• He also said Amazon developed Trainium because customers needed better price performance and because relying on a dominant supplier can leave customers with limited leverage on cost and product priorities.
• The transcript presents strong AI demand as a driver of demand for accelerators, while also pointing to custom chips as an increasingly important alternative.

Takeaways

• NVIDIA is positioned to benefit from strong demand for AI computing, but the discussion also highlights competition from cloud providers’ in-house chips.
• Investors may want to monitor whether custom accelerators such as Trainium become widely adopted and how that affects demand for third-party chips.

Intel (INTC)

• Jassy said Intel had historically held nearly all of the CPU share in a key area of Amazon’s cloud infrastructure. He described Amazon’s development of Graviton as a response to customer demand for better price performance.
• He said Graviton is about 40% more price-performance efficient than comparable x86 processors and is used broadly by Amazon’s largest EC2 customers.

Takeaways

• The discussion illustrates a competitive challenge for established chip suppliers: large cloud customers may develop their own processors when they believe this can improve cost or performance.
• The transcript does not establish the effect on Intel’s overall business, but it points to custom cloud chips as a competitive factor to follow.

Anthropic (Private company)

• Jassy called Anthropic an impressive company and said Amazon has a deep partnership with it. He cited Anthropic’s use of AWS Bedrock and its work with Amazon’s Trainium chips.
• The interviewer cited an estimated $18 billion Amazon investment, but Jassy did not confirm the amount.

Takeaways

• Anthropic is a private-company exposure to consider only where available through appropriate private-market channels; it is not a publicly traded stock.
• The discussion suggests its relationship with Amazon includes cloud access and chip collaboration, but does not provide a valuation, investment terms, or a return outlook.

OpenAI (Private company)

• Jassy said Amazon customers wanted to use OpenAI through AWS and described a partnership that enables OpenAI models to run in Bedrock. He also mentioned a jointly developed stateful runtime environment for building AI agents.
• The interviewer cited an estimated $50 billion Amazon investment, but Jassy did not confirm the amount.

Takeaways

• OpenAI is a private company, not a publicly traded stock. The transcript highlights its commercial relationship with AWS, but does not provide verified investment terms or a basis for valuing the company.
• For Amazon investors, the partnership may help make AWS more useful to customers; the discussion does not quantify its financial contribution.

AI Infrastructure and Cloud Computing

• Jassy described AI demand as concentrated at two ends: large AI labs and a small number of fast-growing applications on one end, and enterprises using AI mainly for productivity and cost reduction on the other.
• He expects a much broader wave of AI use in existing enterprise software over time, particularly as inference costs fall. He said those costs have declined but remain roughly an order of magnitude too high, in his view, for AI to be used as pervasively as possible.
• Amazon faces trade-offs over allocating scarce memory and other components between AI workloads and its established cloud business. Jassy said Amazon intends to keep serving existing customers rather than displace them to prioritize customers willing to pay more for AI capacity.

Takeaways

• The longer-term opportunity depends not only on AI demand from labs, but also on whether enterprises can use AI affordably in everyday software and business processes.
• Track inference costs, enterprise adoption, infrastructure availability, and the returns on data-center investment. The transcript describes strong demand, but does not establish how quickly broad adoption or profitability will follow.

Cybersecurity

• Jassy identified cybersecurity as an immediate concern as AI models find vulnerabilities that might otherwise go undetected. He said companies may need to identify and patch vulnerabilities in minutes, rather than relying on slower, human-led processes.
• He described Amazon’s Continuum service as a tool that identifies vulnerabilities and builds patches, and said AWS also helps customers address security issues.

Takeaways

• The discussion points to potential demand for security tools that can find vulnerabilities and help organizations respond quickly.
• The transcript does not name other cybersecurity companies or provide market-size estimates. A relevant area to monitor is whether businesses can safely automate security testing and patch deployment at scale.

Robotics, Satellite Broadband, and Autonomous Driving

• Jassy identified AI, robotics, and space development as major areas of transformation. Amazon is investing in warehouse robotics, Amazon Leo satellite broadband, and Zoox autonomous ride-hailing.
• He said Amazon Leo is intended to provide broadband connectivity to roughly 500 million households worldwide that lack access.

Takeaways

• These are long-term, higher-uncertainty opportunities within Amazon rather than standalone publicly traded investments.
• For Amazon, the key question is whether these projects develop into durable businesses that justify continued investment; the transcript provides no specific launch dates, financial targets, or expected returns.