AI news, 14 August: Databricks soars to $190 billion valuation

Databricks raises $5 billion at a $190 billion valuation. The data and AI software company said on Thursday it had closed a massive funding round led by existing backers including Coatue, Blackstone, MGX and T. Rowe Price accounts, with new investor Sixth Street Growth, according to Reuters. The valuation is up sharply from about $134 billion just six months earlier. Databricks reported an annualized revenue run-rate above $7 billion and more than 80% year-over-year growth in the second quarter, remaining cash-flow positive on an adjusted basis. Proceeds will support products that help businesses build and run AI applications. The round underscores continued strong investor appetite for companies enabling corporate AI adoption, even as markets watch infrastructure spending closely.

Google launches Gemini 3.7 Flash for coding and agents. Alphabet’s Google released Gemini 3.7 Flash, a lower-cost model aimed at software coding and automated multi-step business tasks, Reuters reported. The company is pitching it for systems that plan work, use tools and complete workflows with less human oversight. Introductory pricing is set at half the prior Flash rate through year-end, and the model is rolling out to Google’s Gemini Spark agent service. Google still has not given a firm date for its delayed flagship Gemini 3.5 Pro model, which investors see as a key test of whether DeepMind can keep pace with OpenAI and Anthropic amid recent leadership changes.

DeepSeek formally releases its V4 Pro model. Chinese startup DeepSeek made its V4 Pro official on Thursday, emphasizing stronger “agent” abilities that let the system handle more complex, multi-step tasks, according to Reuters. The model is available via API, app and web. DeepSeek is also raising API prices and adding peak/off-peak rates. The launch comes as the company expands hiring, computing capacity and fundraising after its earlier low-cost models drew global attention, while facing tougher competition from other Chinese labs.

OpenAI debuts “Ultrafast” mode; IBM deepens enterprise tie-up. OpenAI introduced a preview of Ultrafast, a mode it says lets its GPT-5.6 Sol model run up to 14 times faster and produce as many as 750 output tokens per second, TechCrunch reported. The speed boost is powered by a partnership with chipmaker Cerebras and is initially limited to a small group of customers for uses such as customer support and market analysis. Separately, IBM announced a partnership with OpenAI to bring the lab’s models and tools to more corporate clients through IBM Consulting, including training tens of thousands of consultants and integrating models into IBM’s platforms—expanding OpenAI’s push into large enterprises.

China’s SMIC raises prices on surging AI chip demand. Semiconductor Manufacturing International Corp, China’s leading chip foundry, said AI-related orders continue to drive business and that it has raised prices for high-demand capacity, Reuters reported on Friday. Second-quarter revenue topped $3 billion for the first time and profit roughly tripled, beating estimates. Wafer shipments rose 14% and average selling prices climbed 5.7%. SMIC expects AI demand to stay strong in the second half as it adjusts capacity and ramps new lines, highlighting how the AI boom is tightening semiconductor supply even outside the best-known GPU makers.