Voice takes centre stage in AI race while OpenAI loses battle for Windsurf: this week's news roundup

From manoeuvres for advantage in voice and audio to the antics at Windsurf, here are the stories that grabbed our attention this week.

Voice takes centre stage in AI race

In an apt demonstration of the importance of speech in the race for AI domination, this week has seen the release of Voxtral from French AI startup Mistral and the acquisition of PlayAI by Meta.

PlayAI specialises in human‑sounding voice synthesis, and is likely to be used in Meta’s AI characters, and in wearables like Ray‑Ban smart glasses and Quest headsets. AInvest has analysed the Meta acquisition, saying that “Play AI's core technology – low-latency text-to-speech, multilingual voice generation, and real-time voice APIs – directly addresses Meta's ambition to build "natural-human-level" digital voices”.

PlayAI’s talent is equally critical to Meta’s ambitions, the AInvest report says, calling the acquisition a bold move that gives Meta the edge of Google and Apple, because of its vertical integration of voice AI with AR/VR hardware.

Voxtral from Mistral is a different beast – it’s an open-source audio AI toolkit. It handles up to 40 minutes of audio, including transcription, summarisation, Q&A, and even voice‑triggered API calls, all in multiple languages, and is aimed at businesses and developers. Mistral calls voice “our most natural form of human-computer interaction”. Current voice systems are generally too limited and unreliable for real-world use, it says, but Voxtral will start to bridge this gap.

In short then, Voxtral should speed up innovation in speech AI, while Meta’s PlayAI acquisition signals that voice is now central to how the big players intend to interact with users across social, immersive, and wearable platforms.

Amazon to make further investment in Anthropic

Amazon is reportedly considering further multi-billion dollar investment in Anthropic, in addition to the $8 billion it has committed to put into the AI firm since 2023. Amazon doubled its stake in Anthropic to $8 billion last November, reinforcing its role as Anthropic’s core infrastructure provider. Anthropic’s AI model Claude is integrated into Amazon products and Amazon’s Project Rainier supercomputer programme will be used to build and deploy future versions of Claude.

Amplitude buys Kraftful to speed up AI strategy

Analytics company Amplitude has bought startup Kraftful in a move to accelerate its AI strategy. Kraftful uses proprietary LLM analysis to distill unstructured user feedback data to help teams to identify top feature requests, understand user complaints, and uncover bugs. Kraftful’s capabilities will be integrated directly into the Amplitude platform.

Amplitude says that the integration of Kraftful means that its users will be able to see the full picture of what their customers do and say. A company statement said: “For example, Amplitude will be able to automatically analyse customer feedback and identify a rise in complaints around your checkout flow. From there, you’ll be able to watch a session replay of that experience to identify how you can improve the experience, A/B test the fix, analyse the results, and engage users directly with an in-app notification. All of this will be able to happen within the Amplitude platform, so teams can move from problem to solution without missing a beat.”

OpenAI loses battle to buy Windsurf

Another episode in the high-stakes soap opera at OpenAI: after the public benefit corporation storyline that played out earlier this year, in May OpenAI announced a $3 billion acquisition of Windsurf (formerly Codeium), aiming to integrate its AI-assisted coding tool into ChatGPT. The deal finally fell through last week, amid speculation that OpenAI’s ongoing tensions with its biggest investor Microsoft were at least partly to blame.

No sooner was it known that the deal had collapsed than Google announced it had taken a non-exclusive $2.4 billion licence on certain Windsurf technology, and hired Windsurf’s co-founder and CEO Varun Mohan and some other senior employees to boost its Gemini AI project. It’s a move that enables Google to gain access to Windsurf’s agentic integrated development environment (IDE) technology without a full acquisition, and thus avoid regulatory scrutiny.

Why does any of this matter? It’s not just about the relentless pursuit of top AI talent from the big tech companies that we’ve seen in the last few months, though that of course is part of it. Windsurf’s technology represents the cutting edge of agentic software development, the next leap forward in AI productivity tools. Windsurf has one of the first production-ready agentic IDEs, an environment where autonomous AI agents can plan development, navigate and modify large codebases, run, test and improve code, and collaborate – essentially enabling developers to streamline and speed up software creation. It’s foundational to building AI software engineers, a publicly stated goal for the big tech companies investing in AI.

The story’s not done, however. A couple of days after the Google bombshell, AI startup Cognition announced the acquisition of Windsurf. A statement from Windsurf said: “With this acquisition, two world-class teams are joining forces to shape the next era of AI coding — and we’re just getting started.” Cognition is developing Devin, an AI software engineer designed to perform complex development tasks autonomously, and the Windsurf buy gives it the infrastructure, data, and talent to help fast-track Devin from a prototype into a production-ready AI engineer. 

It could well be a pivotal moment in the AI coding storyline, as big tech companies race to build effective agentic developer tools ahead of their rivals. It’s certainly a significant setback for OpenAI, which has lost out on strategic tech and talent as well as infrastructure. If Google and others nose ahead of OpenAI, then its investment from Microsoft starts to look like a constraint rather than a freedom borne out of well-funded scale and reach.