PLUS: NVIDIA's AI hits the streets, and warnings of AI's exuberance
Good morning
OpenAI has inked a massive $38 billion deal with Amazon Web Services, making a clear strategic move to secure immense computing power beyond its deep partnership with Microsoft.
The agreement provides critical infrastructure for OpenAI's next-generation models and operational independence. Does this signal that a multi-cloud strategy is now essential for any major player in the race to build advanced AI?
In today’s Next in AI:
OpenAI's $38B AWS infrastructure deal
NVIDIA's AI takes to the streets
US approves massive AI chip export to UAE
CEOs warn of an AI investment bubble
OpenAI's $38B AWS Bet
Next in AI: OpenAI has signed a massive $38 billion deal with Amazon Web Services to power its AI models. This multi-year agreement marks a major strategic move to secure immense computing power beyond its deep partnership with Microsoft.
Decoded:
The agreement gives OpenAI access to huge AWS compute clusters, including cutting-edge Nvidia GB200 and GB300 GPUs, with all capacity expected to be online by the end of 2026.
This move is a clear strategy for diversifying from Microsoft, which until this year was OpenAI's exclusive cloud provider for its most intensive workloads.
This collaboration builds upon an existing relationship, as OpenAI’s open models are already among the most popular foundation models available on Amazon Bedrock.
Why It Matters: For OpenAI, this deal provides critical infrastructure and operational independence needed to train its next generation of models. For the industry, it intensifies the cloud war and shows that a multi-cloud strategy is becoming essential for leading AI developers.
NVIDIA's AI Hits the Streets
Overview of the Dublin region data visualisation platform developed in partnership with Bentley Systems, using VivaCity transport data, powered by NVIDIA
Next in AI: NVIDIA and its partners are deploying “Physical AI” in cities from Dublin to Raleigh, using computer vision and digital twins to manage traffic, improve public safety, and optimize urban infrastructure.
Decoded:
In Vietnam, Linker Vision’s platform is being deployed in Ho Chi Minh City after helping cut incident response times by up to 80% in Kaohsiung City, Taiwan.
In Dublin, AI-powered sensors from VivaCity are providing highly accurate data on cyclists, vehicles, and pedestrians to help identify dangerous sites and improve traffic flow.
Milestone Systems is integrating generative AI into its video management platform, aiming to reduce operator alarm fatigue by up to 30% by automating video review and filtering false alerts.
Why It Matters: This shows AI's shift from analyzing digital data to actively managing physical environments in real-time. For urban areas, this means faster emergency responses, safer streets, and more efficient infrastructure powered by live data.
The Great Chip Export
Next in AI: The U.S. government has approved Microsoft's license to export tens of thousands of advanced NVIDIA AI chips to the United Arab Emirates, marking a significant development in the global race for AI supremacy.
Decoded:
The license allows Microsoft to ship 60,400 of NVIDIA's A100 chips, a massive transfer of high-performance computing power.
The agreement also includes NVIDIA’s next-generation GB300 graphics processing units, ensuring the UAE gains access to top-tier AI hardware.
This news immediately impacted the market, causing shares of NVIDIA to climb 2.2% as investors reacted to the deal.
Why It Matters: This move signals a major strategic play in the worldwide distribution of critical AI infrastructure. It also highlights how government approvals are becoming a key factor in shaping national AI capabilities and the competitive tech landscape.
AI's 'Irrational Exuberance'
Next in AI: Top financial CEOs are sounding the alarm on an AI investment bubble, flagging a major disconnect between the massive capital being poured into infrastructure and the actual revenue being generated today.
Decoded:
Big Tech firms like Alphabet, Meta, Microsoft, and Amazon are collectively guiding for more than $380 billion in capital expenditures this year alone.
The caution comes from a belief that significant productivity benefits and revenue gains are a five-year trend, not an immediate return, as consumers are not yet ready to pay for the full cost.
Looking ahead, keeping up with compute demand will require an immense $5.2 trillion in capital for AI-ready data centers by 2030, according to a McKinsey report.
Why It Matters: This situation mirrors historical tech booms, where foundational technologies require enormous upfront investment before their economic impact is fully realized. For builders and investors, this is a clear signal to balance long-term vision with sustainable short-term financial strategies.
AI Pulse
Palantir forecasted revenue of around $1.33B for the current quarter after beating Q3 estimates, outpacing analyst expectations.
Morgan Stanley estimated that global data center capacity will grow sixfold in the next five years, with the hardware and facilities costing $3 trillion by the end of 2028.
Milestone Systems introduced a VLM-as-a-service for developers, using models post-trained on 75,000 hours of traffic video to power new generative AI features in its video management platform.