Skip to content
Daily BriefNewsDaily Brief

9 AI stories from August 15, 2026: Databricks $190B valuation, Gemini 3.7 Flash half-price, Microsoft Copilot merger, and 6 more

· by Pondero Newsdesk · 9 stories

AI news daily brief: 2026-08-15

Nine stories today: Databricks crossed $7 billion in annualized revenue and raised $5 billion at a $190 billion valuation, and Google launched Gemini 3.7 Flash for coding and agents at half the price of its predecessor. Three product updates cover OpenAI's Google Drive integration and macOS context history, Microsoft's Copilot consolidation plus four feature cuts by August 18, and Pony.ai and Uber's 2,000-robotaxi European expansion. Remaining items cover Dynatrace's $915 million acquisition of Arize, Google's AI watermark toggle, Anthropic's SynthID-Text technical disclosure, and an unconfirmed NVIDIA/OpenAI Ohio data center deal.

Databricks raises $5 billion at $190 billion valuation as AI agent demand pushes ARR past $7 billion

Databricks announced on August 13 that it raised $5 billion in a new strategic round at a $190 billion valuation, its second financing event in 2026. Coatue Management led; Blackstone, MGX, T. Rowe Price, and Sixth Street Growth also participated. The company disclosed crossing $7 billion in annualized revenue with more than 80 percent year-over-year growth in Q2 2026, per the official Databricks press release. Three products get the capital: Lakebase, a serverless Postgres database for AI-agent workloads that already passed $100 million ARR; Genie, the AI coworker; and Unity AI Gateway, a multi-model governance and cost-control layer for enterprises running multiple models across clouds. The Lakehouse data warehousing product separately crossed $1.5 billion ARR with more than 100 percent growth. AT&T, Shell, and Toyota are named as enterprise customers.

Full story: Databricks raises $5B at $190B valuation

Google launches Gemini 3.7 Flash for coding and agents at half the price of its predecessor

Gemini 3.7 Flash arrived August 13. Google called it its most capable Flash-class model to date, optimized for coding and agentic workflows. Introductory pricing is $0.75 per million input tokens and $3.75 per million output tokens, half the original cost of Gemini 3.6 Flash, valid through December 31, 2026, per the Gemini API changelog. The model offers a 1 million token context window with up to 64,000 output tokens per prompt. Google says it outperformed comparable models from Anthropic and OpenAI across nine public benchmarks covering code debugging, issue resolution, and web layout generation; specific benchmark names were not disclosed. Available now in the Gemini API, Gemini Spark, and Google AI Studio. For teams running coding or agentic workloads, the temporary half-price window is the most immediate reason to evaluate it.

Full story: Gemini 3.7 Flash launches at half-price

OpenAI adds Google Drive to ChatGPT Library and launches macOS Computer History for ambient context

Two ChatGPT capabilities went live on August 14. The first: Google Drive in the ChatGPT Library, available to Plus, Pro, Business, and Enterprise subscribers. Users can browse Drive files and folders inside ChatGPT, pull Docs, Sheets, and Slides into any chat via @mentions or the composer, and edit documents without leaving the session. Files pulled from Drive are not re-uploaded to OpenAI servers. The second: macOS Computer History, an opt-in feature in the ChatGPT Mac app that lets ChatGPT and Codex reference past app and web activity for persistent context across sessions. It records interaction events rather than screenshots or audio and excludes private browsing, per 9to5Mac. Both features target Pro, Business, and Enterprise users; Computer History is off by default.

Full story: OpenAI adds Google Drive and macOS Computer History

Microsoft merges consumer Copilot and Microsoft 365 Copilot into one app and kills four AI features by August 18

Microsoft is folding two previously separate Copilot products into a single application called Microsoft Copilot. The August 13 announcement put mobile and web rollout in mid-August, with Windows and Mac apps following mid-September. Four features shut down by August 18: Group Chats, AI-generated podcasts, Copilot Labs experimental features, and Deep Research in consumer Copilot, per Microsoft Learn. The animated mascot Mico is also retired. Professional users on paid Microsoft 365 plans get Researcher as the replacement for Deep Research. Microsoft also confirmed that Copilot Chat, Cowork, and GitHub Copilot are merging into one app this quarter. Organizations that built workflows around consumer Deep Research should audit against the August 18 cutoff; the premium Researcher agent requires a paid Microsoft 365 Copilot seat.

Full story: Microsoft Copilot merger and feature cuts

Dynatrace acquires AI observability leader Arize in $915 million cash-and-stock deal

Dynatrace signed a definitive agreement on August 13 to acquire Arize. The deal values Arize at approximately $915 million in cash and stock, including about $815 million in cash plus replacement equity for employees, per Dynatrace investor relations. Founded in 2020 and backed by Andreessen Horowitz, Arize built a platform for AI evaluation, tracing, and observability spanning pre-deployment experimentation through production LLM and agent monitoring. Both founders join Dynatrace: Jason Lopatecki reports directly to CEO Rick McConnell; Aparna Dhinakaran also joins leadership. The acquisition pairs Dynatrace's APM and infrastructure monitoring with Arize's AI-specific evaluation and agent tracing. For teams choosing an AI observability stack, two previously independent layers will now ship from a single vendor.

Full story: Dynatrace acquires Arize for $915M

Pony.ai and Uber expand partnership to deploy 2,000 robotaxis across Europe and Middle East

A commercial robotaxi service already runs in Zagreb, Croatia under the Pony.ai/Uber partnership. The August 14 expansion agreement extends that model to more than 2,000 Pony.ai robotaxis across multiple cities in Europe and the Middle East, per Uber's investor relations page. The phased deployment covers at least four additional cities; specific locations and timelines will be announced separately. Uber provides the mobility platform and consumer-facing interface; Pony.ai provides Level 4 autonomous driving technology and fleet operations. Pony.ai filed a 6-K with the SEC confirming the details. Central and Eastern Europe, where Zagreb provides an existing regulatory foothold, appears the most likely expansion direction based on current geography.

Google adds a toggle to remove visible watermarks from AI-generated content, with EU countries exempt

Starting August 14, Gemini users can toggle off the visible watermark on AI-generated images, videos, and songs inside the Gemini app and the AI video editor Flow. The toggle is rolling out globally except in countries where visible watermarks are legally required, including EU member states operating under the AI Act transparency provisions that took effect August 2, 2026. The invisible SynthID watermark stays active globally regardless of the visible-watermark setting, and C2PA provenance metadata persists, per TechCrunch. The toggle covers outputs from Nano Banana, Omni, and Lyria models. Any company distributing AI-generated content into EU markets should confirm distribution meets the AI Act labeling requirement before toggling watermarks off in an integrated tool.

Anthropic publishes technical explainer on Claude SynthID-Text watermarking: global deployment, limitations on short text and code

On August 14, Anthropic published a FAQ identifying Claude's text watermarking as a deployment of Google DeepMind's SynthID-Text technique, first described in a peer-reviewed Nature paper in 2024. Anthropic disclosed it is applying the watermark globally rather than scoping it to the EU, stating it does not yet have a reliable technical method to watermark by region. Known limitations documented in the FAQ: detection performs poorly on passages under approximately 800 characters; the watermark thins on factual text; and it performs poorly on code, which must be syntactically exact. The watermark carries no identifying information about the individual user, per Unite.ai and The Decoder. Developers building on Claude whose outputs include code or short factual snippets should treat watermark detection as unreliable for those content types.

NVIDIA and OpenAI near agreement on financing a large AI data center in Ohio, per reports

Multiple reports dated August 14 placed NVIDIA close to finalizing an agreement with OpenAI to finance a large data center project in Ohio. Per GuruFocus citing sourced reporting, NVIDIA's financial guarantee was reduced from an original $250 billion to below $120 billion. The Ohio project would draw on NVIDIA's compute-financing program announced August 10 and backed by Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR. Neither NVIDIA nor OpenAI provided official confirmation as of August 14, and no closing timeline was specified, per GuruFocus. Treat as unconfirmed until either company issues an official statement. If the deal closes as described, it would rank among the largest single AI infrastructure financing arrangements on record.

Sources