Capital & Industrial Strategy

Kuaishou's Kling AI has closed a $2.8 billion fundraise with participation from Alibaba and Tencent, according to Reuters. The scale of the round is notable: at $2.8 billion, it is one of the largest single AI fundraises in the Chinese market and reflects the strategic value both Alibaba and Tencent attach to controlling distribution in the generative video segment. Kling has been positioning itself as a direct competitor to OpenAI's Sora and Runway, and this capital injection, backed by two of China's largest platform operators, gives it a runway to compete on model quality, compute, and — critically — distribution through Kuaishou's existing short-video user base.

The Alibaba involvement is particularly significant given the simultaneous news that Alibaba is banning its own employees from using Anthropic's Claude coding assistant, per Reuters. Taken together, these two moves reveal a coherent Alibaba strategy: invest in and back domestic Chinese AI champions while actively pushing employees toward its own Qwen-based stack and away from US-developed tools. This is both a competitive and a regulatory posture, consistent with Beijing's broader push for domestic AI stack self-sufficiency.

The price per unit of AI compute and inference is drifting lower across the sector, according to Bloomberg. This is being read by some analysts as a warning sign: if the enormous capital expenditure on data centres, GPUs, and networking is predicated on sustained or growing revenue per query, declining unit pricing erodes the model. The Philadelphia Semiconductor Index fell as much as 6% on July 2 after its best-ever quarter with an 88% advance, and Richard Windsor of Radio Free Mobile told Bloomberg that markets have fundamentally mispriced the demand picture for AI compute.

The Financial Times notes that a new AI trade is taking hold beneath the surface of headline index gains — one more selective about which AI-adjacent positions are defensible. Allianz Chief Economist Ludovic Subran, speaking at the Aix-en-Provence Economic Forum, stopped short of calling this a bubble but warned of 'exuberance' in AI productivity expectations and expressed doubt that Europe would capture an 'AI dividend,' per Bloomberg. The Wall Street Journal reports similar cracks appearing in Asian tech-heavy indices, with the AI trade diverging sharply between infrastructure winners and application-layer losers. Subran also flagged emerging-market semiconductor exposure as an increasingly attractive position given sustained hardware demand.

Microsoft and Amazon have both created dedicated units of forward-deployed engineers — staff embedded directly inside enterprise customer organisations to assist with AI implementation — mirroring a model pioneered by Palantir and more recently adopted by pure-play AI companies, according to Bloomberg. This is a structurally significant shift in go-to-market strategy for companies that have historically operated at arm's length from enterprise deployments, relying on system integrators and channel partners.

The strategic logic is straightforward: the primary constraint on AI revenue realisation is no longer model quality or availability — it is enterprise integration complexity, change management, and the gap between what models can do in demos and what they can reliably do in production workflows. By internalising that last-mile deployment capability, Microsoft and Amazon are simultaneously accelerating their own revenue realisation, crowding out system integrators like Accenture and Infosys from the highest-margin advisory work, and building proprietary knowledge of enterprise AI failure modes that will feed back into product development. This is vertical integration of the deployment stack, not just customer service.

French President Macron and Indian Prime Minister Modi are both running active campaigns to attract AI infrastructure investment — data centres, cloud build-out, and semiconductor supply chain positioning — with both leaders personally engaging major tech CEOs, according to CNBC. The timing is not coincidental: as the US tightens export controls and China builds a parallel AI stack, mid-tier economies with genuine scale — France with EU regulatory weight and infrastructure, India with labour, data volume, and market size — are positioning themselves as the viable third nodes in a bifurcating global AI supply chain.

For France, the play is to leverage its position inside the EU single market and the EU AI Act compliance infrastructure to attract hyperscaler data centre investment that needs a European regulatory home. For India, the pitch is sovereign AI capacity combined with the world's largest youth demographic and a government actively subsidising domestic AI stack development. Both represent genuine industrial strategy commitments, not just diplomatic optics. The competition between them for the same pool of hyperscaler CapEx is real, and the outcome will shape where non-US, non-China AI infrastructure is concentrated over the next decade.