Artificial intelligence’s next battle isn’t about consumer features but about the foundations that decide whether the industry can scale.

Investors are pouring hundreds of millions of dollars into startups tackling bottlenecks that could stall growth, including chip dependence, enterprise adoption and soaring energy use. Modular, Distyl AI and Empower Semiconductor all raised money last week, underscoring how capital is shifting toward the hidden foundations of the AI economy.

Modular Tries to Break Nvidia’s AI Software Lock

Modular raised $250 million at a $1.6 billion valuation to take on Nvidia’s dominance in the software that programs AI chips, according to a Wednesday (Sept. 24) press release.

For years, developers building on Nvidia’s CUDA Toolkit had little choice but to run workloads on Nvidia’s graphics processing units (GPUs). The CUDA Toolkit is a software kit that lets developers run programs on GPUs, which can handle thousands of calculations at once. The toolkit has become the most widely used platform for accelerating AI workloads, but because CUDA only works on Nvidia chips, it locked developers into its hardware.

Modular is trying to break that lock with a cross-platform AI operating system that lets the same code run on chips from AMD, Apple and others without rewriting code.

If Modular succeeds, enterprises can avoid being locked into a single vendor and shop for performance or cost across hardware suppliers. That could erode one of Nvidia’s biggest moats, even as the chipmaker controls an estimated 80% of the AI data center market, according to The Wall Street Journal.

Advertisement: Scroll to Continue

Backers in Modular’s third funding round included U.S. Innovative Technology fund, DFJ Growth, Google Ventures, General Catalyst and Greylock Ventures, the release said.

Distyl Manages Enterprise AI

Distyl AI raised $175 million at a $1.8 billion valuation, according to a Tuesday (Sept. 23) press release.

Unlike Modular, which sits in the infrastructure layer, Distyl positions itself in the management and transformation layer, embedding its proprietary Distillery AI agent into Fortune 500 firms to re-architect workflows and help them become AI-native enterprises.

Fortune 500 firms increasingly need a system of record for AI, not just a collection of apps. Distyl’s approach appeals to regulated industries such as finance and healthcare, where risk oversight matters as much as productivity gains.

The company will use the capital to expand into these verticals and grow its salesforce, per the release. Lightspeed Venture Partners, Khosla Ventures, DST Global, Coatue and Dell Technologies Capital participated in the round.

Empower Attempts to Cut AI’s Energy Bill

Empower Semiconductor closed a $140 million Series D funding round to scale its power-management architecture, according to a Sept. 22 press release.

AI workloads are driving unprecedented electricity demand. In 2023, data centers consumed more than 4% of the power supply in the United States, according to the Environmental and Energy Study Institute.

Data centers are expected to account for nearly half of U.S. electricity demand growth through 2030, The Wall Street Journal reported in April.

Unlike firms that focus on compute, Empower is selling integrated voltage regulators and vertical power architectures intended to shrink power delivery hardware, cut energy losses and improve throughput in server stacks. For hyperscalers under pressure to expand AI capacity without ballooning energy bills, efficiency solutions like this are as strategic as compute performance.

Investors in the Series D round included Fidelity Management and Research Company, Maverick Silicon, CapitalG, Atreides Management, Socratic Partners, Walden Catalyst Ventures, Knollwood and a wholly owned subsidiary of the Abu Dhabi Investment Authority, the release said.

Different Problems, Same Theme

The three companies address different choke points. Modular tackles hardware lock-in, Distyl supports enterprise adoption, and Empower solves energy efficiency. However, each company addresses a structural barrier that could slow AI adoption if left unaddressed. Investors are putting money into these areas because the challenges are immediate and unavoidable.

The real cost of AI deployment often lies in infrastructure, compliance and integration rather than model fees. Executives remain cautious about scaling more advanced agentic systems until issues of trust and security are resolved. At the same time, analysts warn that compute demand is set to outpace available infrastructure, leaving an $800 billion gap in needed investment.

Together, these pressures explain why investors are steering capital toward companies working on the less visible, but equally critical, foundations of AI adoption.

For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.