Former Swiggy and Zomato executives found AI startup Profound and raised $1.5 million

The tech startup Genius AI has raised $44 million in investment to capitalize on the growing demand in the personal interaction-focused service economy. With this Series D funding round, the company's valuation reached $1.15 billion.

Investment and Partners

Venture firm Lux Capital led the round. Other investors included Bessemer Venture Partners, Imaginary Ventures, L Catterton Growth, 2048 Ventures, and StepStone Private Ventures. These funds will allow Genius AI to shift its focus from beauty booking software to autonomous artificial intelligence tools for a wider range of in-person services.

Company Focus Shift

The company, formerly known as GlossGenius, was founded in 2015 by identical sisters Danielle and Leah Cohen-Shohet. They initially created GlossGenius to help salon owners, aestheticians, and spa operators manage their daily operations. Danielle independently learned programming and wrote the initial code, and over time, the software became a dominant force in the market, helping small businesses process payments, book clients, and promote their services.

Expansion of Genius AI's Mission

The company is now changing its name to Genius AI, signifying a broader mission. As Genius AI, the platform will support all in-person professionals. It is planned for use in medical spas, physical therapy centers, fitness studios, injections, and health clinics.

Autonomous AI for Business

The physical service economy in the United States includes about eight million businesses generating over $2 trillion in annual consumer spending. Although studies show that business owners spend nearly 36% of their working time on administrative tasks, Genius AI aims to completely eliminate this bureaucratic burden. The platform goes beyond simple assistance tools and transforms into an autonomous AI agent capable of handling complex tasks end-to-end, freeing up business owners to interact with clients.

Growth and Healthcare Adoption

The company demonstrates impressive growth and operational scale. Currently, the system is used by over 125,000 businesses across half of the US zip codes. Genius AI reports that its platform is approaching $200 million in annual revenue. To date, the company's tools have helped clients save over 112 million hours of administrative work. To deepen penetration into the healthcare sector, the company has added full HIPAA compliance, allowing physical therapists and clinical specialists to securely store sensitive patient data. This gives Genius AI a significant advantage over traditional tools lacking privacy protection. As artificial intelligence automates office work, human care becomes even more valuable, as skilled professionals like masseuses and trainers provide care that computer code cannot replicate. The new funding is planned to be used to expand software development and increase the engineering team.

Zhipu AI is implementing a multi-level infrastructure development strategy aimed at achieving complete computational independence from the Nvidia ecosystem. According to reports from Bloomberg, Zhipu has completed the construction of a 1 GW data center that operates entirely on domestic AI chips, and some of its operations have already been launched. This center is primarily used to support the development of the GLM series models.

Infrastructure Scale and Transition to Operator

The 1 GW capacity is equivalent to the consumption of approximately 750,000 households, making this data center one of the largest built by an independent Chinese AI laboratory. This step marks a strategic transition for Zhipu from a consumer of computing resources to an owner of infrastructure and a computation operator.

Software Acquisition and Optimization

In addition to the physical base, Zhipu has invested hundreds of millions of yuan in acquiring Zhongke Jiahe. This company develops software for heterogeneous computing and originates from the Compiler Laboratory of the CAS Institute of Computing Technology. Zhongke Jiahe's core competencies include runtimes, virtual instruction sets, inference engines, and compilers. This directly addresses the critical intermediate layer problem between various domestic AI chips and the GLM model architecture. The acquisition allows Zhipu to optimize model performance within the fragmented ecosystem of domestic chips, translating theoretical hardware power into effective computing resources through compiler-level optimization, rather than manual tuning for each chip.

Development of Proprietary Chips

Furthermore, Zhipu is exploring the possibility of creating its own AI chips. According to The Information, preliminary discussions are underway with domestic companies regarding chip design. Although this stage is in the early assessment phase, chip research represents the third and deepest level of the strategy. Instead of endlessly adapting the GLM model to existing hardware, Zhipu aims to create chips specifically optimized for the characteristics of its own model's workloads. This vertically integrated approach—model-software-hardware—resembles the strategies of Google TPU and Amazon Trainium, but it is being applied within the constraints of the domestic semiconductor supply chain.

Strategy Drivers and Market Trends

This push into infrastructure is driven by an acute need for computing power. Sales of GLM Coding Plan were limited in January 2026, and GLM-5 further increased the load on capacities. The self-built data center ensures stable computational capacity that is not dependent on cloud service limitations. Zhongke Jiahe's compiler technology promises acceleration of 50–100 times in certain scenarios.

Zhipu's strategy is similar to the actions of Moonshot AI, which has also established computation operation teams for its own data center and hired chip design specialists to study chips optimized for inference. The market trend is clear: independent model developers are transitioning from the status of renters to owners of computing infrastructure. They realize that at the scale of trillions of parameters, the cost and availability of computation are existential competitive variables that cannot be outsourced. Zhipu's three-tiered strategy—data center ownership, hardware abstraction at the compiler level, and defining proprietary chips—allows it to compete based on the economic efficiency of computation, not just model architecture, potentially creating a barrier to vertical integration inaccessible to pure models reliant on cloud rentals.