The AWS Squeeze

Amazon's Bad Week

For the longest time, the cloud computing business was mainly a two-horse race between Amazon and Microsoft, and everyone else was just fighting for scraps. Amazon Web Services, the division that literally invented the industry, was the undisputed king—the default choice for startups, the dominant platform for enterprises, and the profit engine that funded the entire Amazon empire.

Now, suddenly, its cloud kingdom is under siege. The past week delivered a brutal one-two punch for AWS, a company unaccustomed to being on the back foot. First came the embarrassing internal failure: a massive, 15-hour outage in its most important cluster of data centers. That was quickly followed by the stinging external defeat: Google confirming it had struck a massive deal to supply its custom AI chips to Anthropic—the very AI startup that Amazon has poured billions into to make it the crown jewel of its own cloud.

It's the perfect snapshot of the two-front war AWS is now fighting. On one side, it's battling the internal challenges of its own massive scale—the complexity and "bloat" that can lead to embarrassing outages. On the other, it's facing a newly energized set of external competitors who are using the AI boom to rewrite the rules of the game.

As Matt Day at Bloomberg argued, the perception is that AWS is losing its edge:

Three years into an AI boom spawned by OpenAI's revolutionary ChatGPT, AWS is widely perceived as trailing its tech peers in AI. While AWS remains the cloud market leader, Microsoft Corp. is now growing its backlog of corporate sales faster than Amazon, a trend that recently favored AWS.

Last year, according to a Gartner Inc. estimate, Amazon's cloud division captured 38% of corporate spending on cloud infrastructure services. That sounds hefty until you consider that Amazon's cloud division held almost half of that market as recently as 2018, according to the firm.

So what happened? The game changed. The AI revolution didn't just create a new market; it created a new kind of customer with a new set of demands that AWS's old playbook wasn't perfectly designed to meet.

First, the nature of the AI customer changed the game. AWS' business model is optimized for standardized services and predictable returns. But training a giant AI model requires a completely different kind of approach: a massive, custom-built factory for a single, high-risk "whale" of a tenant. This created an opening for a company like Oracle, which, after being dismissed as a cloud also-ran, was willing to pivot its entire business model to cater to the unique demands of customers like OpenAI.

Second, AI runs best on specialized hardware. This created the opening for Google. While AWS was busy building the world's best general-purpose cloud, Google was perfecting its own custom AI chips, the Tensor Processing Units (TPUs). The Anthropic deal is the ultimate validation of that strategy. It's a signal that the most valuable AI workloads are now up for grabs, and customers are making specific, strategic choices about the underlying silicon.

The result is that the simple two-horse race has turned into a battle royale. As one analyst explained to Bloomberg:

Dave McCarthy, who advises corporate technology buyers at IDC, said clients once mostly wanted to understand the difference between Amazon's and Microsoft's cloud offerings. These days, they're just as likely to ask about Google, Oracle or an upstart like CoreWeave Inc. "There's more choice out there," McCarthy said. "It doesn't bode well for Amazon. It's creating some new competitive pressure that they didn't have before."

For AWS, the innovator that once blindsided the entire tech industry, this is a strange and uncomfortable new reality. The company that built an empire by breaking down the data center into simple, pay-as-you-go pieces is now struggling to compete in an era defined by massive, bespoke, capital-intensive partnerships. The disruptor, it seems, is at risk of being disrupted.

But of course, Amazon is not standing still. For our premium subscribers, we're breaking down AWS's playbook for fighting a war on two fronts.

Extended Analysis: The Amazon Counterattack

On Our Radar

Key signals we are tracking on ARPU Intelligence Desk

Quantum's Classical Brain

• The Headline: IBM has successfully run its critical quantum error correction algorithm in real-time on a conventional AMD chip, a major practical step toward building a commercially viable, fault-tolerant quantum computer. (Reuters)
• ARPU's Take: This is a huge win for both companies. For IBM, it's a massive practical validation of its quantum architecture, proving a critical piece of its roadmap a year ahead of schedule. For AMD, it's a strategic beachhead into the future of computing, positioning its powerful, off-the-shelf hardware as a key enabling component for the quantum revolution.
• The Implication: This breakthrough dramatically de-risks the path to building a useful quantum computer. It proves that the classical "control plane" needed to manage fragile qubits can be built with existing, affordable hardware, not exotic, custom-made silicon. This shifts the narrative for the entire industry from a science-fiction problem to a more tractable—albeit incredibly complex—engineering challenge, potentially accelerating investment and development timelines.

Intel's Turnaround Gets Real

• The Headline: Intel provided an upbeat forecast and beat third-quarter estimates, driven by a rebound in the PC market and buoyed by a series of major investments from the U.S. government, Nvidia, and SoftBank. (CNBC)
• ARPU's Take: This is a critical "proof of life" quarter for Intel. The combination of a recovering PC market and a massive cash infusion from new investors is finally stabilizing its core business. For CEO Lip-Bu Tan, this is invaluable: it provides the financial breathing room and market credibility he desperately needs to execute the much harder, long-term turnaround of the company's manufacturing and foundry businesses.
• The Implication: This signals a shift in the market's perception of Intel from a company in freefall to one with a viable, albeit long, path to recovery. Investors are now willing to look past the immediate manufacturing gap with TSMC and focus on the stabilizing fundamentals. This buys Intel a precious window of time to prove its long-term strategy can work.

P.S. Got a specific intelligence challenge not covered here? Drop us an email at arpu@hedder.com—we occasionally explore ideas to inform our core coverage.