The final week of February brought a flood of AI model launches, hardware deals and policy moves. The figure that grabbed Wall Street’s attention came from Google, Amazon, Meta and Microsoft, which together pledged roughly $650 billion in 2026 AI infrastructure spending across data centers, custom silicon and cloud expansion. That spending wave sharpened the market split over whether AI is becoming a durable productivity engine or a capital-heavy trade running ahead of returns.
Major model releases dominated the week
Google Deepmind introduced Gemini 3.1 Pro on Feb. 19, highlighting stronger reasoning and a 1 million-token context window. The company said the model posted benchmark gains and expanded multimodal performance, allowing longer sessions across text, code and images.
Anthropic’s Claude Sonnet 4.6 also drew heavy attention. The release improved coding and long-context reasoning while keeping prior pricing in place. Anthropic also rolled out Claude Cowork, a desktop AI agent built to interact with local files and browsers, reflecting the fast rise of agent-style AI products.
In China, Alibaba’s Qwen 3.5 stood out for scale, with 397 billion parameters and a mixture-of-experts design aimed at cost efficiency. The report said its open-weight approach points to a push for broader enterprise adoption in robotics and manufacturing. Bytedance, for its part, released Seedance 2.0, a generative video model that can produce realistic clips from text, images or existing footage, while adding tighter safeguards after earlier controversy around synthetic media misuse.
Spain-based Multiverse Computing added Hypernova 60B to the mix. Built with quantum-inspired compression techniques, the model is being offered free through developer platforms and Hugging Face, with the promise of lower inference costs for coding and tool-calling workloads.
Hardware spending reset market expectations
The model race was only part of the story. Infrastructure spending made an even bigger impact. Alongside the four-company $650 billion commitment, OpenAI was reported to have signed a $10 billion agreement with Cerebras Systems for wafer-scale chips delivering hundreds of megawatts of compute capacity. The plan is tied to faster inference for products such as ChatGPT and support for more complex models through 2028.
Edge AI also moved into focus. Ambiq expanded research operations in Singapore to push ultra-low-power edge computing for wearables and industrial systems. Another notable capital move involved xAI, the Elon Musk-founded company behind Grok. The report said a large Saudi-linked investment flowed into the firm, showing how sovereign money is entering the AI buildout.
Regulators moved at the same time
Governments were active as product releases accelerated. In the United Kingdom, officials expanded plans for free AI skills training for 10 million adults by 2030 and advanced guidance for AI-ready datasets. In the European Union, policymakers published a draft transparency code under the AI Act, spelling out labeling requirements for generated content and clarifying rules for high-risk systems.
The pace matters. Regulation is no longer being discussed in the abstract; it is starting to take shape in specific compliance requirements.
AI adoption spread into everyday operations
The week also showed AI moving beyond demos and into workflows. Reuters reported measurable newsroom gains, saying AI tools helped cut corrections by 10% while supporting journalists with data analysis, though human editors remain in charge.
In biotech, software firm Benchling said its latest industry findings showed 73% adoption of AI tools in protein prediction. Retailer Lowe’s rolled out AI voice agents nationwide to manage customer calls, allowing staff to focus on in-store assistance. Samsung also partnered with Gracenote to improve smart TV search and recommendation through AI-driven metadata analysis.
Those developments gave both sides of the market more evidence to point to. Bulls focused on automation, reasoning engines and efficiency gains. Bears kept watching rising capital expenditures and rich valuations, questioning how quickly monetization can catch up.