Hard to keep pace with the capital pouring into the platforms. TL:DR: billions for Anthropic, Thinking Machines, Perplexity and more. Meanwhile, governments are treating AI as defence infrastructure: the Pentagon is directly funding labs, while China is pouring billions into compute and open models. The AI boom is as much geopolitics as venture.
IPOs / Public
Via (on-demand transit software) filed for IPO after raising $1B and expanding to 650 cities in 30 countries. Figma may make more noise, but Via could mark a steady unicorn debut this quarter.
Meta is refusing to sign the EU’s voluntary AI code, calling it regulatory overreach just weeks before binding rules take effect – setting up a confrontation between Brussels and Silicon Valley.
Meta hired Mark Lee and Tom Gunter from Apple’s AI team, weeks after poaching their boss Ruoming Pang with a $200M package. Confirms the fierce Big Tech talent war and philosophical divergence – Apple’s “tools” vs Meta’s “agents”.
Big Dogs
OpenAI is testing a checkout feature inside ChatGPT, aiming to take a cut of purchases and convert free users into a revenue stream – a shot at Google’s dominance in shopping search.
Thinking Machines Lab (AI lab founded by Mira Murati) raised $2B at a $12B valuation despite no public product, pitching OSS tools for researchers and model builders.
Lovable (AI app builder) became Europe’s latest unicorn – just 8 months after launch, it raised $200M Series A led by Accel at a $1.8B valuation.
Perplexity (AI-powered search/browser) is now valued at $18B after letting investors buy in at a premium just two months after closing a $14B round.
Windsurf (AI coding startup) saw its $3B sale to OpenAI collapse; instead, Google hired its founders and licensed tech in a $2.4B “reverse acquihire”. Days later Cognition stepped in to acquire the remainder, salvaging GTM value and employee payouts. Highlights intensifying AI talent wars and creative M&A to dodge antitrust.
Investors are considering $100B for Anthropic. NB Given how quickly they grew from $1bn-> $4bn revenue, $100 billion might sound cheap – but AI model shops burn a LOT of cash. That said, margins are getting better (as are OpenAI’s) so economics are improving - nascent operating leverage. Nonetheless, the venture industry is incredibly levered against future AI success – the question “what if the curve lags expectations” is increasingly top of mind.
Venture Capital Insights
US startups raised $162.8B in H1 2025 (+75.6% YoY), mostly AI-driven. In contrast, VC firms raised just $26.6B (–33% YoY). Founders with AI leverage dominate, everyone else sidelined.
Redstone launched an ELTIF fund offering professional investors fractional ownership in 200+ European tech startups, reflecting EU reforms (ELTIF 2.0, MiCA) and the push to democratise private equity access. Startups are staying private longer, public markets lag innovation.
Fintech has raised <10% of global VC this year – its weakest share in over a decade.
A “red-hot consolidation wave” is underway – if you’re not exit-prepped, you’re missing the play. Windsurf’s back-to-back weekend acquisitions perhaps underline the point! (Venture Insider)
Venture Geopolitics
The Pentagon awarded $200M contracts each to Anthropic, Google, and xAI (after OpenAI last month) to advance agentic AI for battlefield use. Locks talent and compute into US defence, with strict demands for safety and auditability.
China is pouring billions into AI via infrastructure, subsidies, and open models tailored to censorship rules. Firms like DeepSeek and Alibaba are closing the performance gap with US rivals.
China approved Synopsys’s $35B takeover of Ansys, creating a US duopoly that now controls nearly the entire market for advanced chip design software (EDA). This is a critical chokepoint in the semiconductor supply chain – without EDA tools, no cutting-edge chips can be designed. With both firms American, Washington effectively controls a global bottleneck, while Beijing must either build domestic EDA capabilities or accept dependence on US-controlled software. The move underscores how M&A and antitrust in one sector translate directly into national security leverage.
Trump-backed crypto bills (GENIUS and CLARITY) failed to advance in the House after a surprise procedural vote, denting “Crypto Week” momentum and hitting stocks like Circle and Coinbase.
Strategic sectors
AI
AI safety researchers at OpenAI and Anthropic criticised xAI’s “reckless” safety culture, citing inadequate guardrails.
Silicon Sands warned of an “AI metrics crisis”: benchmarks like MMLU are outdated, misaligning expectations with real-world performance in customer service, content, and analysis. The result is disappointment and waning confidence.
Blogs by Kiro.dev and sjtdigital.com highlighted how quickly incumbents can release features that make AI startups redundant, echoing earlier notes on the startup–incumbent race.
Cybersecurity
Exein, a European start up that embeds cybersecurity directly into chipsets to protect connected devices / edge infrastructure raised an $81.2M round at a $290.2M, led by Balderton.
Heka Global a startup that helps banks and FIs detect and prevent digital fraud using AI that monitors transactions and behaviour in real-time raised $14M from Windare, Barclays.
Defence
Both Washington and Beijing now treat AI as a national security asset – the US by direct Pentagon funding, China via state-backed infrastructure and OSS models. The AI arms race has blurred the line between venture funding and defence policy.
XTEND “AI-powered autonomous robotic systems for defence and emergency response” raised $30M from Aliya and Protego.
Crypto
Zerohash which helps financial institutions offer crypto trading, stablecoin transfers, and digital asset rewards by handling the backend infrastructure is reportedly in the market to raise ~$100M at $1B valuation (Fortune).