Andy Jassy

a16z · Amazon CEO Andy Jassy on the $220 Billion AI Buildout

• Why Did Amazon Rebuild Alexa?

  Andy Jassy says Amazon had to rewire Alexa because generative AI made the old brains too limited to stay relevant. Those brains were inventive when Alexa was built, but the possible capabilities were so different that starting over was obvious, and still difficult, because the teams were proud of what they had built.

  He says a company that treats AI as a small upgrade to a successful product will very likely not have the winning experience three to five years from now. Retail, Prime Video, Fire TV, and Prime Music are less obvious than Alexa, and he says Amazon has teams reinventing those too.

  ▶ at 48:02 Share Transcript

• Why Did Amazon Buy Annapurna Labs?

  Andy Jassy says Amazon did not buy Annapurna Labs in 2015 for AI chips. He calls the Annapurna team one of the most talented he has worked with, and the deal one of the best acquisitions in Amazon's history, if not the best.

  Amazon had been working with Annapurna on a network card that offloaded virtualization, networking, security, and storage so nearly 100 percent of an EC2 server could go to customer compute. That card ended up in every server. Andy Jassy says Intel then had virtually all of the CPU share and was not driving price performance, so the Annapurna team built Graviton. He says 99 percent of Amazon's top 1,000 EC2 customers use Graviton, and it delivers about 40 percent more p

  ▶ at 27:58 Share Transcript

• Do Golf Courses Use More Water Than Amazon Data Centers?

  Andy Jassy says golf courses consume 200 times more water than all of Amazon's data centers, and more water in two days than those data centers use in a year. He puts almond farming in the same category.

  He calls that water comparison one of the myths aimed at data centers, which he says have become an easy villain. On pollution, he says backup generators are registered as if they ran full time, but they run about 10 hours a year for maintenance.

  ▶ at 18:01 Share Transcript

• How Fast Is AWS Growing?

  Andy Jassy says AWS grew 36.7 percent year over year last quarter, on a $169 billion annual revenue run rate.

  He says that business could be a few hundred billion dollars of revenue over about 10 years, and that AI means customer experiences will be reinvented with inference. He also points to Prime Video as a large and profitable business, the low-Earth-orbit satellite constellation Amazon Leo, autonomous ride-hailing with Zoox, healthcare, and robotics.

  ▶ at 4:24 Share Transcript

• How Much Has Amazon Grown Since 2021?

  Andy Jassy tells Ben Horowitz and Erik Torenberg that compared with 2021, when he took over as CEO, Amazon's overall revenue is about 80 percent higher, AWS is almost three times bigger, and ads revenue is about two and a half times bigger.

  On the retail side, he says Amazon doubled a fulfillment network that had taken 25 years to build, and did it in about 18 months, then built a last-mile transportation network the size of UPS.

  ▶ at 1:43 Share Transcript

• How Does AI Change How Companies Fix Security Flaws?

  Andy Jassy says companies will have to rewrite almost all of their code so agents can update pieces of it under real constraints, and so security fixes can ship in minutes.

  The nervousness of the last month or two about models, he says, has focused on fantastical scenarios: an evil botnet taking over everything, or robots infused with malicious intent. He thinks the industry and the public and private segment will figure the fantastical cases out together. The risk he says gets too little attention is cybersecurity. The models find far more vulnerabilities than humans were finding. Applications will have to respond to imminent security threats in minutes. Today, people classify vulnerabilit

  ▶ at 41:46 Share Transcript

• Why Did Amazon Invest in Anthropic and OpenAI?

  Andy Jassy says Amazon's investments in Anthropic and OpenAI come down to who the teams are, the technology and the models they have built, some of the applications they have built, and Amazon's belief in how transformative AI is. He calls both companies really impressive teams and really impressive companies.

  The question put the checks at about $18 billion in Anthropic and $50 billion in OpenAI, and the asker said he was estimating. It also said similar investments from NVIDIA had gotten Wall Street up in a tizzy because they recalled the dot-com boom. Andy Jassy says the rate at which the two companies have grown, at least on the top line, is really unbelievable, and he does not know if

  ▶ at 33:45 Share Transcript

• How Does Amazon Split Memory Between AI and Core Customers?

  Andy Jassy says Amazon has to split scarce memory between HBM for AI accelerators and DRAM for core, non-AI customers. He says 2026 capacity is sold out, 2027 is largely sold out, and a good chunk of 2028 is sold out. He does not think the shortage of data center space, power, and memory ends in 2027.

  AWS leaves some capacity for on-demand use in its inference service, Bedrock, or EC2, but the big chunks are largely sold out. The AI business and the core business are both growing at rapid rates.

  Andy Jassy says the starting point is being customer-focused. In the pandemic, he thinks competitors tried to optimize within quarters and drive as much revenue as possible while companies were cut

  ▶ at 24:26 Share Transcript

• Why Is $220 Billion of Amazon CapEx Still Not Enough?

  Andy Jassy says Amazon expects to spend about $220 billion of CapEx in 2026, and even at that level the company is leaving a meaningful amount of demand on the table. He says no equation makes the size of the buildout simple.

  In the first three years of the generative AI wave, the AI business in AWS is over a $25 billion run rate. After three years, early AWS was about a $60 million annual run rate business, so the AI business is several hundred times faster. He remembers a leadership breakfast 15 years ago when a telco spending $50 billion of CapEx sounded unbelievable.

  In the early 2000s, product teams spent 70 to 80 percent of their time on infrastructure that did not differentiate them

  ▶ at 10:40 Share Transcript

• How Fast Did Amazon Build Fulfillment?

  Andy Jassy says Amazon decided to build out its fulfillment network and last mile transportation network aggressively and quickly in the early part of the pandemic. He says going that fast would be really hard, would probably cost more than going slower, and would probably not get everything as optimally as a slower build. No physical stores were open, customers and sellers needed the capacity, and Amazon decided, as it almost always does, to shade on the side of customers.

  Not everything in the plan went perfectly. A lot of the fulfillment team was in the office more than most of the rest of the company, while many people designing, planning, and assessing the network were remote and sprea

  ▶ at 6:40 Share Transcript