DeepSeek Reopens $8 Billion Round at $74 Billion Valuation-Monolith Money Meets an AI Power Hogs' Bill

DeepSeek's reopened raise puts Monolith alongside a major China AI bet

DeepSeek has resumed a round for close to $8 billion at a valuation near 500 billion yuan, or about $74 billion, after pausing the process last month. Bloomberg also reports that Monolith Management is in talks to contribute, which would sit oddly but interestingly alongside reports that Monolith is raising capital for a separate Chinese AI vehicle. Either way, the reopening suggests investors still see DeepSeek as a meaningful China AI name.

Part of the raise is expected to support data-center buildout, including a 1-gigawatt AI data center in Inner Mongolia. That makes the raise more than a model-finance story: it is also a bet on compute capacity and infrastructure in a market where capital is still chasing the next bottleneck.

Cheaper models are widening AI demand, not capping it

Lower training costs changed the economics

Before DeepSeek disrupted expectations, training a frontier model could cost over $100 million. DeepSeek's breakthrough came at under $6 million, a shift that lowered a major barrier to entry and made advanced AI more accessible to a broader set of buyers and builders.

DeepSeek's release also triggered talk of a DeepSeek death zone for rivals, reflecting how much pressure the company put on both price and performance. The key point is not that cheaper AI reduces infrastructure demand. It is that lower costs can broaden adoption and keep demand for compute, power, and related infrastructure firm.

Commercialization appears to be sustaining the buildout

Early after DeepSeek's release, some market moves suggested investors worried that cheaper AI would cool the spending cycle. In January, Nvidia lost nearly $600 billion in market cap as those concerns flashed through the market. But major companies later recommitted to major investments in new data centers, suggesting the slowdown narrative was premature.

Commercial use cases also started to widen quickly. By April 2025, DeepSeek had announced partnerships with BMW and BYD for in-car AI and with XCMG for industrial modeling, according to market-intelligence coverage of the company's early adoption curve. That helps explain why a cheaper model can still support heavier infrastructure demand: more applications and more production usage can offset the benefit of lower per-task costs.

DeepSeek's planned price hike is the next real test

After the funding restart and the attention around adoption, the next question is whether DeepSeek can convert scale into better economics. The company told users it plans substantial price increases across its AI services. That matters because it tests whether low-cost dominance can turn into real pricing power rather than just higher usage.

The debate is straightforward. DeepSeek's listed pricing is still far below many rivals, so there is room to raise revenue per request without immediately breaking demand. At the same time, open-weights access and competing hosts could limit how much extra revenue the company actually keeps. For investors, that is the core watchpoint: whether higher prices stick, or whether usage shifts elsewhere because the underlying model is easy to replicate.