Two stories last week underscored how venture geopolitics frames the AI era. First, Coatue argued that AGI could help stabilise US debt-to-GDP around 100% by 2034 (and even reduce it to ~80% under ASI), a provocative claim that highlights how investors increasingly see intelligence as a macroeconomic variable on par with capital and labour. Second, the sheer scale of corporate AI capex – $370bn+ by 2026 from Google, Amazon, Meta, and Microsoft – shows how tech giants are now national champions, at once advancing US strategic aims and presenting themselves as neutral partners abroad. Together these stories illustrate how capital, productivity, and power are converging in AI – a shift shaping both markets and states. Outside of venture geopolitics, Figmas IPO deserves some recognition.
IPOs / Public
Figma IPO: Priced at $33, hit $112 before trading was halted, valuing at more than $50bn (over 2x Adobe’s $20bn offer in 2022, blocked by UK regulators). Stock was around 40x oversubscribed. Q2 profits were $9–12m on $247–250m revenue, with 40% YoY growth, 88% gross margin, 13M+ MAU, and 132% NDR.
It was the largest first-day pop for a US IPO in more than 30 years, reigniting IPO pricing complaints. Some argue this derisks the IPO pathway for others like Circle, CoreWeave, and Chime.
The Figma success also led Dealroom to ask “is Index the new Sequoia?” Index’s 66m shares are worth more than $2.2bn ($6.6bn including the uplift) – potentially a triple fund returner even for a fund of Index’s scale.
Index’s 2025 so far also includes Google’s $32bn acquisition of Wiz, Meta’s $29bn partial exit of Scale AI, and Revolut hitting a $75bn valuation.
Klarna is ready to IPO as early as September.
Firefly Aerospace (space tech) set the IPO mid-point at around $5.8bn – it will set the tone for future space tech exits, including SpaceX.
AWS, Microsoft/Azure, and Google all reported earnings. Meanwhile, Alphabet, Microsoft, Amazon, and Meta are set to spend nearly $400bn on capex this year.
Google is rolling out Gemini 2.5 Deep Think, its most advanced reasoning model, able to answer questions considering multiple ideas simultaneously, priced at $250/month. AI overview summaries, integrated in late 2024, have radically altered behaviour – click-through rates to websites dropped nearly 50% according to Pew. This is good for users but catastrophic for the digital marketing ecosystem, with desktop CTRs for publishers down 48% YoY. Google still reported strong ad revenue.
Microsoft is on track to become the second company after Nvidia to reach a $4trn market cap. Quarterly profits soared, with net income up 24% YoY to $27.2bn.
Zuck noted that “in the last few months,” Meta has “begun to see glimpses of our AI systems improving themselves.” Q2 net income reached $13.5bn, up 36%. The company lifted its forecast, sending its share price soaring. Integration of AI into its advertising is already generating “meaningful” revenue, and engineers will soon be allowed to use AI during coding interviews.
Amazon recorded net income of $18.2bn for the quarter through June, up 34%.
Tether reported net profit of $4.9bn in Q2, breaking the $4.5bn record set in Q1. Global supply of USDT is around $163bn. Circle’s USDC, the next most-owned stablecoin, has a global supply of $64bn. Though Tether remains one of the world’s largest holders of US government debt, its customer base is almost entirely international.
Reddit is investing heavily to become a “go-to search engine.” Shares rallied after it crushed earnings expectations.
Palantir reported a 53% YoY revenue increase from the US government to $426m, and commercial revenues also almost doubled. The CEO cited “astonishing impact of AI” driving results. Palantir announced it had signed a deal with the US government worth $10bn over the next decade. Shares have more than doubled since the end of last year, with a market cap now at $379bn.
Big Dogs
OpenAI raised $8.3bn from a mix of new investors (including $2.8bn from Dragoneer) and existing backers, as part of its previously announced $40bn raise, valuing the company at $300bn. ARR has surged to $12–13bn. OpenAI is in complex negotiations with Microsoft to restructure its non-profit governance, seen as a prerequisite to going public. The current contract runs until 2030, and under the new deal Microsoft may receive about one-third equity stake. GPT-5 is set to launch this month. OpenAI also introduced “study mode,” which avoids giving direct answers to students and instead asks open-ended, Socratic-style questions.
Anthropic’s latest valuation is $170bn, tripling in just five months, with revenues climbing from $4bn to $5bn run rate.
Manus (China) is building a platform to orchestrate teams of AI agents for complex research and reasoning tasks. It hopes to outmanoeuvre larger Western players by breaking down hard problems into auditable, multi-agent workflows.
Lovable Dragos Novac shared an optimistic take: $100m ARR in eight months is an early indicator that we’re entering a new economic era, with vibe coding at its core. His case:
It is here to stay – fast, low-friction, intuitive, AI-native.
It has never been easier to create digital products, mirroring the explosion of social content five to ten years ago.
Exponential adoption is inevitable, not just because more people are developing, but because they are running multiple projects in parallel.
This compounding effect is driven by developer-centric models, LLM-native tooling, and ambient AI assistance – Lovable is positioned at the centre.
Compared to Replit, which took three years to reach $100m ARR, Lovable’s speed is unmatched. Its current $1.8bn valuation at a 24x multiple looks conservative versus category leaders like OpenAI (25x), Anthropic (40x), and Replit (30x). Projections suggest $1.2–3bn ARR by 2027 and $5.2bn ARR by 2029. Novac suggests Lovable could be Europe’s engine for a new AI-native developer economy.It is the eighth European unicorn of 2025 (others include MUBI, Neko, Quantum Systems, Parloa, Tines, Sygnum, and Loft Orbital).
N8n is in talks with multiple VCs to raise fresh funds, just four months after a €55m Series B. The number of term sheets is now into double figures, according to multiple sources, with valuations north of $2bn.
Venture Capital Insights
Andreessen Horowitz noted that for 15 years, consumer software meant free with ads or perhaps $20/month. Now prices cluster around $200/month – 10x the old ceiling. With this new pricing, you only need 41,000 customers to build a $100m ARR company, which changes everything about building consumer software.
Coatue keynote argued that AGI could stabilise the US debt-to-GDP ratio at around 100% by 2034 (and with ASI, potentially fall to 80%), far below current projections of ~120–140%. Their forecast assumes a causal chain – more intelligence leads to more productivity, which lifts GDP and eases the debt burden. The Economist published a counter-argument questioning whether such productivity assumptions are realistic.
Venture Geopolitics
Google, Amazon, Meta, and Microsoft plan to spend more than $370bn by 2026 on AI (data centres, talent, infrastructure). That is equivalent to the GDP of the Czech Republic and reflects how AI strategy has become deeply geopolitical. In the US, these firms are framed as national AI champions under the White House’s AI Action Plan. Abroad, they present themselves as apolitical global partners supporting local AI ambitions. This balancing act means they are simultaneously advancing US interests and scaling globally.
The Economist on the economics of superintelligence: before 1700, the global economy grew at just +8% per century. Industrialisation lifted this to 350% over 300 years. AI faces no such demographic limits. Once machines perform 30% of tasks, annual GDP growth could exceed 20% (the only ceiling being the laws of physics). Labour markets would be reshaped as wages are capped by the cost of compute, with elite human talent and AI-relevant capital owners capturing most wealth. Others would be squeezed by collapsing prices for AI-generated goods and inflationary “cost disease” in human-dependent sectors. Financial markets could swing wildly as investors chase winners in a winner-takes-all world, potentially driving interest rates to 20–30%. Politics would become volatile with vast inequality, new tax regimes, and civil rights risks. “Humanity may soon be outpaced in intelligence – but will need wisdom more than ever.”
South East Asia has become a key front in the global AI and tech race. The region has 700m people and 2GW of data centre capacity – equal to London and Frankfurt combined. Singapore hosts 60% of regional capacity and balances geopolitical ties deftly, awarding tenders equally to US firms (Microsoft, Equinix) and Chinese firms (GDS, Bytedance). Many Chinese firms (e.g. PC Partner, Manus) are relocating HQs to Singapore to distance themselves from Beijing. Bytedance insists it is not a Chinese firm, running global ops from Singapore. Others bypass US chip bans by renting GPU servers in SE Asia – legal but controversial. Smuggling persists: Singapore police arrested suspects accused of moving $390m of banned Nvidia chips into Malaysia via Singapore. As legal and illegal supply chains blur, SE Asia is becoming platform, buffer, and back door in the tech economy – strategically vital and geopolitically sensitive.
In the UK, Coinbase released a satirical musical ad “Everything Is Fine” mocking Britain’s economic and social climate while pitching cryptocurrency as an alternative to a failing system. Regulators banned it. A YouGov poll showed that 40% of Brits never use AI and only 10% use it daily. Those who do mostly seek “simple topic” answers. The UK is trying to age-gate swathes of the internet. Google is already experimenting with an “age estimation model” to detect whether users are under 18 (using signals like YouTube viewing history). UK Research & Innovation announced a $74m programme to attract world-class researchers.
Strategic Sectors
Cybersecurity
Palo Alto (market cap $136B, $2.3B cash, no long-term debt) announced the $25bn (19x) acquisition of CyberArk, adding identity as its fourth core platform (alongside SASE, SecOps, and Cloud). The move signals Palo Alto’s ambition to dominate in identity. It throws down the gauntlet to peers like CrowdStrike, Fortinet, Check Point, and Zscaler, who will need to respond or risk irrelevance as identity converges with security. Standalone identity vendors such as Okta, SailPoint, Ping, and Saviynt face existential pressure to either bulk up their platforms or find acquirers. Palo Alto has also announced a bold target of $15B next-gen security ARR by 2030 (vs $5B in Q3 2025), which requires 20% CAGR. Just days earlier, there were rumours Palo Alto was considering acquiring SentinelOne ($1bn ARR in endpoint security).
Energy
Trump’s AI Action Plan flagged the US energy crunch as a threat to “AI dominance.” Hyperscalers are responding with huge capex: Google with Kairos Power (nuclear), Amazon with X-energy (nuclear), Google and Meta with geothermal, Microsoft with hydrogen. Flexibility is also critical: xAI has joined programmes to avoid grid use at peak times, gaining priority access to power. Abroad, the Gulf and Spain are booming with data-centre expansion. Michael Liebreich’s model shows that as long as clean energy outgrows total energy demand by a few percentage points annually, fossil fuels will inevitably be squeezed out. China’s nuclear strategy illustrates the compounding: reactors now cost $2–3/W, far less than the US’s Vogtle 3 and 4 ($15/W). China could overtake the US in nuclear capacity by the early 2030s. Meanwhile, “greenhushing” is rising: despite political pressure in the US leading companies to downplay climate messaging, most are actually strengthening their commitments, including supply chain emissions.
Robotics
SemiAnalysis mapped a five-stage framework of robotics progress. We are now at Levels 2–3, where robots navigate messy environments and perform some low-skill tasks. As robots gain dexterity and tactile intelligence, they will step into Level 4, enabling automation of tasks once thought untouchable: skilled trades, fine manufacturing, and even caregiving.
Defence
Frontier labs like Microsoft, Anthropic, and Mistral are partnering with national security to deploy models (ClaudeGov, Mistral’s Saba) on classified data. Operational use is slow, limited by tooling, reliability, and legacy processes. Experts warn the real challenge is re-engineering mission workflows, not just model deployment. Concern is growing that while the US may lead in developing AGI, China could still lead in applying it effectively, extracting faster insights without the same guardrails. The risk: the US wins the AGI race but loses on real-world adoption.