画像クレジット: Image via Source article. Used under fair use for news commentary. · source
A single week in June 2026 saw roughly $65 billion in AI financing activity announced across two deals. That number is worth sitting with for a moment. It represents more capital than the entire global robotics industry raised in 2024, concentrated into just two transactions involving a Chinese startup and a consortium of American private equity giants.
The first deal involves Moonshot AI, a Beijing-based company seeking up to $2 billion at a $30 billion valuation. This would mark the startup's third financing round in six months, a cadence that suggests either extraordinary momentum or an extraordinarily capital-intensive burn rate (or, to be precise, probably both). The second involves Apollo and Blackstone teaming up on a $35 billion infrastructure financing arrangement that would fund AI chip procurement for companies including Anthropic and Broadcom. Together, these deals represent something genuinely new: not just large sums of money, but potentially new categories of AI investment entirely.
I should note upfront that we're working with limited information here. Bloomberg broke both stories, but the details on terms, conditions, and even whether these deals will close remain sparse. What follows is analysis based on what's public, with appropriate uncertainty where the reporting is thin.
Moonshot AI's trajectory is remarkable by any standard. A $30 billion valuation would place it among the most valuable AI startups globally, though the company's actual products and revenue remain somewhat opaque to Western observers. What we do know is that the Chinese AI landscape has become intensely competitive since the release of DeepSeek's models earlier this year, and that competition appears to be driving a funding arms race. Three rounds in six months is not normal venture capital pacing. It suggests either that Moonshot is growing so fast that each round genuinely reflects new value creation, or that the company needs continuous capital infusions to maintain its position against well-funded rivals. The distinction matters enormously for how we should interpret this valuation, but I haven't found public data that would let us adjudicate between these explanations.
関連記事
More in AI Models
One uses graph-based reasoning to auto-generate rewards; the other fuses human language and physical corrections. Both beat expert-designed baselines.
James Chen · 8 hours ago · 5 min
Three new papers tackle the same problem: how do you get a robot to understand 'I left my backpack on the table' when it can't even see the table?
Sarah Williams · 9 hours ago · 4 min
Two new papers tackle the unsexy problem that's actually holding back robotics: we can't generate enough good training data without armies of human experts.
Mark Kowalski · 11 hours ago · 6 min
The collaboration hints at where large enterprises are placing their bets on AI automation, though the technical details remain frustratingly sparse.
The Apollo-Blackstone deal is, in some ways, more interesting from a structural perspective. Private equity firms creating dedicated financing vehicles for AI chip procurement represents a departure from traditional technology funding models. Venture capital funds startups. Growth equity funds scaling companies. But who funds the physical infrastructure that AI requires? Until now, the answer has mostly been the hyperscalers (Microsoft, Google, Amazon) and the chip companies themselves. The emergence of third-party infrastructure financing suggests that demand for compute has outstripped what existing capital structures can provide.
To be precise about what's actually novel here: the innovation isn't that private equity is investing in technology (that's been happening for decades), but that Wall Street is creating what Bloomberg describes as "an entirely new AI investment category" specifically for chip financing. The $35 billion figure is striking, but the mechanism might matter more than the amount. If this model works, it could allow AI companies to access compute without the dilution that comes with equity financing, and it could allow infrastructure investors to participate in AI's growth without betting on which specific models or applications will win.
There are reasons to be skeptical, though. I know I'm being picky here, but the economics of chip financing are genuinely tricky. AI accelerators depreciate rapidly as new generations arrive. NVIDIA's H100s, which commanded premium pricing eighteen months ago, are already being displaced by H200s and B100s. A $35 billion bet on chips is, in part, a bet that those chips will retain enough value over the financing period to generate returns. That's not obviously true, and I haven't seen analysis of how Apollo and Blackstone are structuring the deal to manage depreciation risk.
The involvement of Anthropic in the Blackstone-Apollo deal adds another dimension. Anthropic has been notably aggressive about compute acquisition, reportedly spending billions on NVIDIA chips to train its Claude models. Access to a dedicated financing facility could accelerate that spending further, potentially widening the gap between frontier labs and smaller competitors. Whether that's good for AI development broadly is a question worth asking, though it's probably too early to say.
What connects these two deals is a shared underlying reality: AI development is becoming extraordinarily capital-intensive, and the capital requirements are growing faster than traditional funding mechanisms can accommodate. Moonshot's rapid-fire fundraising and the creation of bespoke chip financing vehicles are both responses to the same pressure. The question is whether this capital intensity is a temporary phase (reflecting the cost of building foundational capabilities that will eventually become cheaper) or a permanent feature of the AI landscape.
The optimistic read is that we're in an infrastructure-building moment analogous to the railroad era or the early internet. Massive upfront capital expenditure creates the foundation for decades of subsequent innovation. The less optimistic read is that AI is becoming a game that only the very well-capitalized can play, with implications for competition, innovation, and the distribution of AI's benefits.
It's worth noting that these deals are happening against a backdrop of genuine uncertainty about AI's near-term economics. The major AI labs have not, to my knowledge, demonstrated that their products generate revenue commensurate with their costs. That doesn't mean they won't, but it does mean that $65 billion in new financing is, in some sense, a bet that the economic model will eventually work out. The investors involved are sophisticated enough to understand this, which suggests they see something in the trajectory that justifies the risk. But sophisticated investors have been wrong before, sometimes spectacularly.
For the robotics and embodied AI space specifically, these financing trends have mixed implications. On one hand, abundant capital for AI infrastructure could accelerate the development of models that enable more capable robots. The foundation models that power humanoid reasoning, manipulation planning, and environmental understanding all require the kind of compute that these deals are designed to provide. On the other hand, if capital continues to concentrate in a small number of frontier labs and infrastructure plays, smaller robotics-focused AI companies may find themselves squeezed out of the compute market entirely.
There's also a geopolitical dimension that deserves attention, though I'll admit my expertise here is limited. Moonshot AI's rapid valuation growth is happening in China, while the Apollo-Blackstone deal is structured around American companies and chips. The parallel development of massive AI financing capacity in both countries could reflect a kind of capital arms race, with each side seeking to ensure it has the resources to compete in what both governments clearly view as a strategically important technology. Whether this competition accelerates AI development or distorts it in problematic ways is genuinely unclear.
What I'd want to see next is more transparency on several fronts. For Moonshot, understanding what products and capabilities are driving this valuation would help assess whether it's grounded in fundamentals or primarily reflects competitive dynamics. For the Apollo-Blackstone deal, details on the financing terms, depreciation assumptions, and risk-sharing arrangements would help evaluate whether this model is replicable or a one-off. And for the AI industry broadly, better data on the relationship between compute spending and capability improvements would help us understand whether the current capital intensity is buying proportional advances or hitting diminishing returns.
For now, we're left with a striking data point: in a single week, the AI industry attracted financing commitments that exceed the GDP of many countries. That fact alone doesn't tell us whether we're witnessing the birth of a transformative industry or the inflation of a spectacular bubble. Probably, as is usually the case with complex economic phenomena, the truth involves elements of both. The capital is real, the technology is advancing, and the uncertainty about where it all leads remains profound.
