Not long ago, the PGA Tour shipped its own servers on trucks around the country. That means every week, employees loaded up racks of computer equipment and scoreboards, hauled it to golf courses in new cities, unloaded, rigged it together, and ran the tournament’s data operation before tearing it down and rebuilding it again on another course.
“The PGA Tour is a traveling circus,” said David Provan, the tour’s vice president of digital architecture. “We don’t play inside a set of white lines.”
The golf tour’s old setup is a small-scale illustration of a pattern that Amazon CEO Andy Jassy predicts will shift in increasingly large numbers over the next two decades. Jassy told investors on Thursday’s second-quarter earnings call that about 85% of global IT spending is currently for on-premise company servers, an equation that “is going to flip in the next 10 to 20 years,” he said.
Amazon’s Q2 results, fueled by a soaring growth from Amazon Web Services, show the acceleration is happening faster than expected. AWS revenue hit $42.2 billion in the second quarter, a 37.6% jump from a year earlier, the division’s fastest growth rate in 18 quarters, and the fifth consecutive quarter of acceleration. AWS added more than $4.6 billion in revenue sequentially—about 80% more than its previous largest quarterly increase ever—and is now running at a $169 billion annualized clip. Customer backlog stands at $496 billion, and Amazon’s stock is surging on the results, up more than 15% on Friday.
The PGA Tour, which is sponsored by AWS and signed a new agreement during Amazon’s first quarter, illustrates the mechanism behind the massive acceleration. Companies are moving their existing infrastructure into the cloud, and then from there, stacking their AI on top of it. The result is a flywheel that explains AWS’ staggering growth and why Jassy told analysts on Thursday that he now believes AWS will “very possibly be a $1 trillion annual revenue business for us in time.”
Investors are eating it up, despite Jassy telling the world Thursday that Amazon’s capital expenditures will actually be $220 billion in 2026, up from its previous estimate of $200 billion.
“AWS growth continues to astound,” William Blair analysts wrote in a note published on Friday. “With Amazon’s track record, high AWS margins, and accelerating growth, we believe the market is likely to react more favorably to the raised capex guide than feared, especially as Jassy touts the potential for AWS to be a $1 trillion revenue business with time.”
‘Nine Minutes’
When the PGA Tour opened new global headquarters about five years ago, it didn’t include fixed IT infrastructure, said Provan, and instead moved everything over to AWS. Its “ShotLink” ball-tracking system on the course, a media archive, the website, mobile app, and data distribution systems all made the migration, he said.
With the data now moved to the cloud, Provan said the tour went on to layer AI on top of the data and applications. It’s a pattern Amazon executives say is becoming common.
“Increasingly, customers seeking the full benefits of AI are accelerating their transition to the cloud,” Amazon CFO Brian Olsavsky said Thursday. “As customers invest in AI, we see a corresponding increase in core consumption.”
The PGA tour now uses an agentic AI platform on Amazon Bedrock—AWS’ service for deploying foundation models—to generate betting profiles for each week’s tournament field. It can create about 150 profiles in “about nine minutes,” Provan said. Through Bedrock, the tour also produces thousands of pieces of content each week, and at the 2025 Players Championship, the group launched AI-generated commentary. The feature offers a two-sentence summary for each of the 32,000 shots hit during a tournament.
In 2026, Provan said the tour debuted what he called “agentic production” which is a fully automated television broadcast with camera selection and graphics driven by ShotLink data. The Tour also built a “favorite player hub” in its app so when golfers like Scottie Scheffler finish a round, the app’s homepage will show AI-generated insights about how he played.
“People follow players more than they follow tournaments,” Provan said. “They follow their human.”
But none of it began as an AI-based playbook, noted Provan. When AI first gained popularity, Provan said the tour focused on “AI projects” it wanted to execute. It has since shifted to testing how AI can help solve business problems. The internal mantra now is that the company doesn’t have AI projects, but it has projects that use AI.
“And it’s OK if everything doesn’t use AI,” he added.
Jassy’s barbell
This week, Jassy framed the current state of AI demand as a barbell. On one end are the AI labs like Anthropic and OpenAI that consume enormous amounts of computing power, along with a few breakout consumer applications like Claude Code and ChatGPT. On the other end are enterprises that generate value from AI through productivity and lower costs gained via automating customer service, streamlining business processes, and catching fraud.
In the middle of the barbell, Jassy said, are all the existing enterprise production workloads that could use AI inference but mostly aren’t yet. That middle segment, he told analysts, “will be the largest absolute segment” over time and is “going to change very significantly.”
“We’re still in the relative early stages of how much demand there’s going to be for AI,” said Jassy. “I think it’s going to change every customer experience that we know.”
Val Henderson, CEO of AWS-exclusive consulting firm Caylent, sees the same trajectory playing out across plenty of companies. She said Bedrock’s model-choice architecture, which Jassy touted on Thursday, is the single most compelling differentiator AWS offers. The ability to run Anthropic, Meta, OpenAI, and Amazon’s Nova on one platform with prompt routing to the most cost-effective model is one of AWS’ main advantages, she said. For CEOs, CFOs, and boards, clarity in return on investment has been a critical piece.
“Speed is just one element,” said Henderson. “You have to think about accuracy, [and] you have to think about cost.”
Amazon disclosed that customers spent more on Bedrock in the second quarter than in every prior quarter combined since the service launched in 2023. More customers have joined Bedrock in the past six months than the first two years, and AWS’ AI business has exceeded a $25 billion annual revenue run rate.
Matt Wood, AWS’s chief AI and technology officer, said in a recent interview that the enthusiasm he sees from customers has not peaked the way he once expected. AI remains at the top of every customer’s agenda, he said, and the conversation has broadened well beyond technology leaders to include chief human resources officers, chief financial officers, and boards of directors. What surprised him, he said, is which industries are moving fastest.
“It’s actually the regulated businesses,” Wood said, pointing to healthcare, financial services, insurance, and manufacturing. He attributed the pace to the fact that the regulatory environments those companies operate in have forced them to focus on governance and data security for years, and that data turns out to be a prerequisite for deploying AI effectively. The incremental investment to layer AI on top, he said, is relatively small.
Wood compared the role of pricing in cloud computing to a restaurant; low costs were the initial draw, but the capabilities that kept customers coming back were the real value.
‘Cost was often the sizzle that got people into the restaurant,” he said. “But the steak was the agility and the transformational nature of having access to all these new capabilities.”
For Provan, the PGA Tour arc from trucks packed with servers and automated broadcasts has been like a long car ride with passengers helping navigate, he said.
“The tour has really gone through an entire technological transformation,” he said. “And AWS has ridden shotgun with us the whole way through it.”
