TheChek News Flow

Shake-up follows fears tech giant missed early AI boom as Microsoft and Google challenge cloud business

Amazon is embarking on the largest capital spending programme in its history, seeking to regain momentum against AI rivals by expanding data centres, developing chips and building models.

The group is undergoing a strategic shake-up amid fears its cloud arm, AWS, is losing ground to competitors in securing corporate AI contracts, according to more than a dozen current and former senior employees.

Chief executive Andy Jassy last week announced Amazon’s capital expenditure would rise to $200bn this year, exceeding that of Google and Microsoft, with the outlay focused on computing infrastructure.

In December, he consolidated the group’s chip, model and advanced research teams under a single leadership structure, a move intended to align its AI plans. The Amazon chief has also cut costs, including jobs — eliminating some 30,000 of about 350,000 corporate roles.

“We have deep experience understanding demand signals in the AWS business and then turning that capacity into a strong return on invested capital,” Jassy said earlier this month. “We’re confident this will be the case here as well.”

AWS employees said the company’s moves also reflected internal concern that it had failed fully to capitalise on its lead in cloud computing, in particular by being slower than rivals to secure major contracts with AI providers after OpenAI launched ChatGPT in 2022.

“We were just not fully prepared for how fast things would unfold,” said one former senior AWS employee.

AWS remains the world’s largest cloud provider, generating nearly $130bn in sales last year and more than 60 per cent of Amazon’s overall profits. But analysts forecast that demand for AI-powered cloud services will lead Microsoft’s cloud business to overtake AWS over the next three years.

Amazon said other providers did not report “true cloud figures” and that this prevented an accurate comparison between different cloud businesses.

Roughly three-quarters of Amazon’s planned $200bn capital expenditure is allocated to AWS, according to public filings. Microsoft, Google and Oracle are on course to collectively spend nearly $400bn this year.

Jassy last week said Amazon was planning to add a meaningful amount of data centre capacity this year. In 2025, it added nearly 4 gigawatts of capacity — equivalent to the annual energy consumption of more than 3.2mn US homes. The group plans to double capacity by 2027.

Investors are uneasy about the scale of Amazon’s bet. Shares are down more than 20 per cent from their November peak amid concerns about how quickly the spending will translate into returns.

Employees said Amazon was under pressure to secure more deals with leading AI groups, among the main drivers of new cloud business. It has invested $8bn in Anthropic and is building vast data centres for the group. But its initial equity investment came after Google backed the start-up.

Meanwhile, as one of OpenAI’s earliest investors, Microsoft secured exclusive cloud computing contracts with the ChatGPT maker.

Only after Microsoft allowed OpenAI to undergo a corporate restructure did Amazon sign a $38bn cloud computing deal with the start-up last year. That agreement is dwarfed by OpenAI’s $250bn contract with Microsoft and its $300bn worth of deals with Oracle.

Amazon on Friday said: “It’s not accurate to infer that AWS was unable to secure major compute deals or was at a disadvantage when it comes to capacity planning. AWS continues to earn most of the big enterprise and government transitions to cloud.”

It has touted the uptake of the group’s Graviton and Trainium chips, used for conventional cloud computing and AI training respectively. Sales of these chips are on course to generate more than $10bn in combined annual revenue.

Amazon debuted its latest generation of Trainium chips in December, promising a significant increase in performance. It is holding talks to join OpenAI’s latest multibillion-dollar funding round in a move partly designed to ensure the ChatGPT maker adopts its semiconductors, said people familiar with the matter.

The chips should also help Amazon reduce its reliance on Nvidia’s products, helping to expand AWS’s profit margins from renting out data-centre capacity to corporate customers.

However, Google has attracted interest in its “tensor processing unit” (TPUs), using its custom chips to advance its Gemini AI models. The search giant sold Anthropic 1mn TPUs in a deal worth tens of billions of dollars.

Ben Bajarin of tech consultancy Creative Strategies questioned whether leading AI start-ups would adopt Amazon’s chips for their core products even if they proved cheaper to run than Nvidia.

“Amazon talks specifically about price performance but the problem is that some users need outright performance,” he said.

Amazon is also spending to advance its own “Nova” AI models, marketing them as a low-cost alternative to rival models.

Nova underperforms the most advanced models made by OpenAI, Google, Meta and Anthropic, according to independent benchmarks. Executives have been irked by some AWS employees describing Nova as “Amazon Basics”, a term used for the group’s generic household products, according to three people familiar with the matter.

Staff said the company was pushing its own AI tools such as developer platform Kiro, while setting a target for 80 per cent of developers to use AI for coding tasks at least once a week. But several company engineers said they preferred using Anthropic’s Claude over Nova for coding work, with one AWS engineer saying: “I didn’t even know we had a model.”

Amazon said its Nova models were used by “tens of thousands of AWS customers” and performed comparably to some leading models. It noted AI labs such as Anthropic were also deploying a significant number of its Trainium chips.

Jassy last year said: “We’re going to keep pushing to operate like the world’s largest start-up — customer-obsessed, inventive, fast-moving, lean, scrappy, and full of missionaries trying to build something better for customers.”

The pressure to regain ground in the AI race is weighing on employees. Some said they feared Amazon could slip closer to “day two” — a term used by founder Jeff Bezos in 2018 to describe a business in “stasis”, followed by an “excruciating, painful decline”.

“The culture has shifted, but so has the world around us,” said one senior AWS engineer of the company’s strategic pivot. “We’re going to have to prove our worth.”

70 GW of new solar capacity is scheduled to come online in 2026–2027 → a 49% increase in operating solar capacity from the end of 2025.

Airbnb says its custom-built AI agent is now handling roughly a third of its customer support issues in North America, and it’s preparing to roll out the feature globally. If successful, the company believes that in a year’s time, more than 30% of its total customer support tickets will be handled by AI voice and chat in all the languages where it also employs a human customer service agent.

“We think this is going to be massive because not only does this reduce the cost base of Airbnb customer service, but the quality of service is going to be a huge step change,” CEO Brian Chesky said during the company’s fourth-quarter earnings call this week. This seems to suggest he believes the AI would do a better job than its human counterparts in resolving some issues.

The company also touted its recent hire of CTO Ahmad Al-Dahle, poached from Meta for his AI expertise, and its plans to create an AI-native experience.

With his guidance, Chesky said that Airbnb was poised to introduce an app that doesn’t just search for you, but one that “knows you.”

“It will help guests plan their entire trip, help hosts better run their businesses, and help the company operate more efficiently at scale,” Chesky explained, adding that’s why Airbnb brought Al-Dahle on board.

“Ahmad is one of the world’s leading AI experts. He spent 16 years at Apple, and most recently led the generative AI team at Meta that built the Llama models. He’s an expert at pairing massive technical scale with world-class design, which is exactly how we’re going to transform the Airbnb experience,” Chesky noted.

Like other businesses poised for disruption by AI, Airbnb’s leadership is pushing the idea that it has a unique database and product that other AI chatbots can’t replicate.

“A chatbot doesn’t have our 200 million verified identities or our 500 million proprietary reviews, and it can’t message the hosts, which 90% of our guests do,” Chesky told analysts during the earnings call. Instead, he pitched the idea of layering AI over the Airbnb experience, which he claimed would help to accelerate growth.

The company forecast revenue growth would be in the “low double digits” this year, after pulling in $2.78 billion in the fourth quarter, above estimates of $2.72 billion. This quarter, it expects revenue of $2.59 billion to $2.63 billion, above Wall Street forecasts of $2.53 billion.

Investors still wanted to know if AI platforms could be a risk in the long-term, assuming they moved into the short-term rentals market. Chesky, however, pushed back at that idea, saying that Airbnb isn’t just the consumer-facing app; it’s also the host app, the customer service, and the protections it offers, like insurance and user verifications.

“We’ve built this over 18 years. We handle more than $100 billion in payments through the platform,” he said.

Meanwhile, AI chatbots serve a function similar to search, in that they deliver top-of-funnel traffic, he noted. That traffic also converts at a higher rate than traffic from Google, Chesky pointed out, suggesting that the shift to AI would benefit Airbnb.

The company is already using AI to power its search, with the feature now enabled for a “very small percentage” of Airbnb’s traffic, while it experiments with making its search more conversational. Later, the company plans to integrate sponsored listings within search.

While Spotify this week told investors its best developers hadn’t written a single line of code since December, thanks to AI, Airbnb offered a more high-level metric on its own internal AI adoption. The company said that 80% of its engineers now use AI tools, and it’s working to get that to 100% soon.