Deepseek - a SaaS Bypass?

The buzz around Deepseek is that China finally has a ChatGPT contender. But the real earthquake might be happening beneath the surface: China could be poised to leapfrog the entire B2B SaaS model as we know it.

Let's rewind. China has seen spectacular success in consumer technology (B2C). Think of Alibaba and Tencent - massive gravity well super apps anchoring a galaxy of user experiences. Yet, B2B SaaS in China? Let's just say they're still struggling to land on the moon.

Several factors contribute to this – regulatory hurdles, a lingering distrust of cloud services, and a cultural preference for one-off purchases over subscriptions. But I believe the root cause is China's hyper-competitive business environment.

It's no secret that Chinese companies… borrow. Designs, business models, technologies – they're all fair game. What often surprises Western observers is the intensity of internal competition and…borrowing. Chinese startups operate in a Darwinian jungle, and those that survive emerge as global powerhouses (see: Temu, TikTok, and the ongoing electric vehicle reckoning).

This brutal competition breeds a fierce protectiveness of company data. B2B SaaS fundamentally relies on trust: you're handing over your precious data, trusting that the vendor won't misuse it. This trust is… less prevalent in China, which helps explain SaaS's relative lack of traction.

Enter Deepseek and its ilk. Chinese enterprises are now rapidly deploying their own compute infrastructure. For a relatively modest investment (a few hundred thousand USD), a company can run a frontier-level LLM on-premises with a fairly decent rate of tokens per second.

SaaS also relies on standardized, repeatable processes – the ability to sell the same core product to many customers. But what happens when the cost of software development plummets, and local LLMs can be trained on a company's proprietary data? A new business model may be emerging.

Imagine, instead of a standardized product, adaptive solutions tailored to a company's unique processes and data, all powered by their in-house LLM. Initially, human consultants will drive this customization. Gradually, inevitably, the machines will take on more of the load.

So, what comes after Software as a Service? Perhaps a return to first principles: Service as a Service, but redefined. Experts in specific problem domains will still scale their expertise, but instead of selling a pre-packaged product, they'll bring a solution, fine-tune it to the client's workflows, and integrate it with the client's proprietary LLM and data stack.

Human expertise, problem-solving skills, and relationships will become even more critical. Just as China bypassed traditional consumer retail and jumped straight to mobile super-apps, it might be about to leapfrog the Western B2B software market, landing directly in a post-SaaS world.

“The power to destroy a thing is the absolute control over it.”

― Frank Herbert, Dune