TMTPost compares Kimi's product-led path with DeepSeek's infrastructure bet

On October 19, 2026, TMTPost published a report contrasting Kimi and DeepSeek, framing them as two diverging paths: one product-driven, the other infrastructure-driven. According to the report, early 2026 saw a popular narrative that the two companies were converging. Both had completed large funding rounds, both had put long context, Coding and Agent work on their roadmaps, and their job openings overlapped. TMTPost argues the opposite: across funding cadence, revenue structure, technology choices, commercialization posture and organizational form, the two chose different directions at each key juncture.

The report traces Kimi's funding rhythm. Its valuation was about $4.3 billion at the end of 2025. After a roughly $2 billion round completed in May 2026, its post-money valuation passed $20 billion. Another round around the July launch of K3 brought the post-money figure to $35 billion, followed by a final pre-IPO round at a $50 billion pre-money valuation. Kimi then confidentially filed an A1 document with the Hong Kong Stock Exchange. A person close to the deal told the outlet that Kimi prices its valuation off its product cycle, with each model generation serving as a window to raise again.

DeepSeek followed a different cadence. It brought in outside equity financing for the first time in 2026, with a first round of about RMB 51 billion and a planned second round of RMB 50 billion, corresponding to a post-money valuation of about RMB 500 billion, while preparing a listing on the Shanghai Stock Exchange's STAR Market. The report converts that to roughly $75 billion, more than twice Kimi's valuation in the same period. TMTPost notes that DeepSeek's valuation increase does not hinge on a single model launch or a steepening revenue curve; it reads more like a one-time pricing within a relatively closed system. On where the money goes, the report says Kimi puts it into compute and talent density, while DeepSeek puts it into building large clusters of its own. The report stops at this comparison of the two paths and offers no further developments.