AI Trends — 18 August 2026

Two labs, two very different balance sheets

The past 48 hours sharpened a growing contrast between the industry's top two labs. OpenAI is pushing toward a public listing that could value the company near $1 trillion, even as it's projected to lose roughly $30–40 billion in 2026 and isn't expected to turn a profit until 2030. Anthropic, by contrast, is coming off its first operating profit and is now reportedly in talks to buy AI infrastructure startup Decart for about $6 billion — its largest acquisition to date. Decart builds software that helps chips run AI workloads more efficiently, and the deal would help Anthropic stretch its existing compute to meet surging demand ahead of a rumored October IPO of its own.

Google keeps shipping

Google released Gemini 3.7 Flash just three weeks after 3.6 Flash, with meaningful coding and automation gains: FrontierCode 1.1 climbed from 34.4% to 43.6%, DeepSWE v1.1 rose from 49% to 65.3%, and AutomationBench jumped from 17% to 30.4%. Separately, Google is retiring three Imagen 4 model IDs today, directing developers to migrate to gemini-3.1-flash-image, and confirmed it has now passed 1 billion users.

OpenAI pushes speed alongside scale

OpenAI opened a limited API preview of "Ultrafast" mode for GPT-5.6 Sol, powered by Cerebras hardware, delivering roughly 750 output tokens per second — up to 14x the throughput of its standard offering. The move continues a broader industry shift toward optimizing for latency and real-world task completion, not just raw model capability.

The efficiency and safety undercurrent

Beyond the headline deals, researchers reported combining synthetic DNA with a semiconductor to build an ultra-low-power memory device capable of storing and processing information in the same location — part of a wider push toward more efficient AI hardware. On the policy side, a key tranche of the EU AI Act's requirements, covering high-risk systems and GenAI transparency rules, began applying this month, adding new compliance pressure for labs operating in Europe.

The takeaway

The story of the week is diverging fortunes: OpenAI racing toward a trillion-dollar valuation while burning cash, and Anthropic converting profitability into deal-making power. Meanwhile Google keeps iterating fast on both models and infrastructure, and efficiency — in chips, in inference, in compliance — is becoming as important a battleground as raw capability.

Sources:

Compiled from public reporting on AI developments over the past 48 hours.