A lot of the AI narrative has focused on the companies it might destroy. Software firms, the argument goes, are being squeezed as customers bypass apps and go straight to powerful AI from OpenAI, Anthropic, or Google.
But a different story is emerging. Ironically, one of the biggest winners from China’s flood of cheap, open models could be the very software industry that many feared would become obsolete.
Chinese companies, including Moonshot, DeepSeek, and Z.ai, are releasing models that are close to the performance of the best American systems but at a fraction of the price. Many are “open weight,” meaning software companies can download key parts of them, run them on their own infrastructure, customize them for their own products, and avoid relying entirely on expensive outside providers.
That changes the economics of providing software in the AI era. Instead of paying premium prices every time a customer asks an AI-powered question, software companies can choose from dozens of competing models. They can reserve premium, closed American models for the hardest tasks while routing more routine work to cheaper, open alternatives.
Open-weight gains
Developers are already embracing this approach. Vercel CEO Guillermo Rauch says open-weight models accounted for about 55% of tokens flowing through his company’s AI Gateway in July, up from just 4% in January.
The shift accelerated after the releases of DeepSeek V4, Z.ai’s GLM-5.2, and Moonshot’s Kimi K3. OpenAI, Anthropic, and Google still dominate spending, but their combined share has fallen sharply as developers experiment with cheaper alternatives, according to data from Vercel’s AI Gateway.
“Unambiguously good”
That competition is exactly what software companies have been hoping for.
William Blair analysts argued recently that cheaper, more open AI models are “unambiguously good” for software companies because AI is simply another input cost. As those costs fall, profit margins improve, and companies can afford to add AI to many more products.
Software companies also gain more control by mixing and matching models rather than depending on a single supplier, or downloading Chinese model weights and using them as the basis for an entirely new AI model that they own and control. Cursor has done this successfully, and other US companies are trying it, too.
As new, powerful Chinese models appeared in the past month, there’s been a noticeable rebound in enterprise software stocks, after a brutal first half of 2026.
HubSpot has soared almost 30% in the past month, while Adobe is up more than 20%. Intuit, Salesforce, ServiceNow, and Asana have all gained at least 12%.
Instead of worrying that AI will replace SaaS, there’s a growing realization that many software vendors can benefit from cheaper Chinese AI models rather than be displaced by them.
Startups rejoice
That’s why venture capitalists are watching this trend so closely. Former Benchmark partner Bill Gurley argues that startups, cloud providers, chip companies, enterprises, and researchers all benefit when powerful AI models remain open and inexpensive. For startups in particular, free models mean they can build AI products without paying a pricey toll every time a customer uses them, while avoiding dependence on any single AI supplier.
One of Silicon Valley’s less-discussed realities is that many of today’s “AI startups” are not building giant foundation models from scratch. They’re building software that sits on top of existing models. Their competitive advantage comes from the customer experience, their industry knowledge, and the data they combine with AI — not from inventing the underlying intelligence.
Cheap Chinese open-weight models make these businesses dramatically more attractive by lowering one of the highest costs of adding AI features. That’s good news not just for startups, but also for the venture capital firms betting on them. If those models remain freely available, a much larger generation of software companies can afford to compete.
This shows how the balance of power in AI could be shifting. If intelligence becomes cheap and interchangeable, the real value shifts away from the model itself and toward the software built around it. The winners become companies with trusted customer relationships, unique data, and products people already use every day.
For at least a year, software companies worried AI would eat their lunch. Instead, thanks in part to an unexpected wave of cheap Chinese models, they may be getting the biggest discount in the industry’s history.
Sign up for BI’s Tech Memo newsletter here. Reach out to me via email at abarr@businessinsider.com.
