The hard part of corporate AI isn't building a clever model — it's getting one to actually work inside a business. Amazon's cloud division, AWS, is now spending about $1 billion to attack exactly that problem, creating a new unit of "forward-deployed engineers" who embed directly with customers to build and ship their AI systems, CNBC reported.

What AWS is doing

Rather than just selling cloud tools and walking away, AWS will send thousands of its own engineers to work alongside client teams — on-site or closely embedded — co-writing production code, wiring AI into the customer's actual workflows, and training staff to run it afterward, per TechCrunch. It's a high-touch, consulting-like model layered on top of selling cloud capacity.

(Explainer: a "forward-deployed engineer" is a vendor's engineer who works inside a customer's organization to custom-build software, rather than handing over an off-the-shelf product. The approach was popularized by Palantir and is now spreading across AI.)

Why it matters

The move targets a real and growing problem: companies are spending heavily on AI but struggling to deploy it. Many pilots never make it into production — by various industry estimates, the large majority stall — because firms lack the in-house expertise to bridge the gap from demo to working system. AWS is betting that hands-on help is what unlocks that spending.

It's both offense and defense. On offense, embedding engineers drives more cloud consumption and ties customers in tightly. On defense, AWS is responding to rivals: as Boursel has covered amid the broader AI-capex boom, Microsoft, OpenAI and Anthropic have all moved into enterprise-AI services — OpenAI and Anthropic reportedly with their own large, well-funded ventures — so a high-touch offering has become table stakes rather than a differentiator. AWS is still the largest cloud provider (with roughly a 30%-plus share, ahead of Microsoft's Azure, per industry trackers), but enterprise AI is the battleground where that lead is defended or lost.

The bigger picture

The build-out reflects an AI-services land-grab. It's not just the cloud giants: consulting firms like Accenture, Deloitte and McKinsey are racing to sell "applied AI," sensing companies will pay a premium for help turning AI hype into real cost savings and revenue. AWS says its embedded teams have already compressed some deployments from months to days for early customers (a claim from the company; results will vary).

What it means

For AWS, the $1 billion signals that owning the cloud isn't enough — to keep growing, hyperscalers must get deeper inside how customers actually use AI. For enterprises, it's a fix for a genuine pain point: the shortage of talent to move past pilots. And for the AI economy Boursel has tracked — huge spending chasing uncertain returns — it's a notable shift in where the money goes: from raw computing power toward the people and services that make AI pay off. Boursel offers no view on Amazon's shares; the takeaway is that the AI race is increasingly being fought not in the data center, but inside the customer's office.