Analysis of AI compute, model deployment, agent infrastructure, and the technical stack for enterprise technology buyers and operators.
16 insights in this category
Buyer resources
16 results
GPT-5.6 turns model selection into a routing decision: Sol for frontier reasoning, Terra for balanced production, and Luna for high-volume work. The enterprise trade-off sits in price, token efficiency, tool orchestration, safety controls, and vendor dependence.
AI token prices fall tenfold a year, yet enterprise bills keep climbing. Google processes 1.3 quadrillion tokens a month. The pattern has a name — and it dates to 1865.
Microsoft, Meta, OpenAI and Oracle all signed up before the first machine was built. Inside the giant computer AMD made to challenge Nvidia — and what it costs to run.
JPMorgan is running AI that works alone for hours. Three-quarters of companies are still stuck in trials. What separates the two — and what closing the gap actually costs.
Singapore's digital economy hit S$128.1B — 18.6% of GDP. Tech Directory SG's July 2026 report maps the telecom, subsea cable, data centre, AI and semiconductor landscape.
Software that acts on its own now changes records across IT, HR, finance and security. It needs its own login, a full record of what it did, and a way to be stopped.
Three days after Fable 5 reached the public, Anthropic said a U.S. export-control directive forced it and Mythos 5 offline worldwide for every customer.
Singapore has become the place AI companies route through for capital, trust, cloud infrastructure, pilots and Southeast Asian access. That is a powerful position. It is not the same as being a frontier AI lab.
AI models now take in sound, video, sensor readings and even touch — and react in under 200 milliseconds. That is fast enough to run factories, ports and warehouses.
Andrej Karpathy has joined Anthropic's pre-training team, shifting back into frontier-model R&D while stepping away from day-to-day work at Eureka Labs.
AI Engineer Singapore, the first Asia edition of the AI Engineer conference series, brought frontier-lab speakers, four builder tracks, and a cabinet-level keynote into Singapore's civic district.
Model Context Protocol can reduce one kind of integration friction for enterprise agents. It does not remove the need to control identities, tool permissions, data flow, evaluation, and incident response.