Expanse wants to cut wasted GPU capacity. Investors put up USD 5.3m

Expanse, a startup founded by University of Edinburgh graduates, has raised USD 5.3m from international investors. Polish fund AIP Seed also took part in the seed round. The company is developing technology designed to help businesses make better use of the costly computing power needed to develop AI.

The artificial-intelligence boom is driving demand for ever more computing power. The problem is that companies do not always use the resources they already have efficiently. Expanse, a San Francisco-based startup co-founded by Polish entrepreneur Nikodem Bieniek, wants to build a business around that gap. The company has just raised USD 5.3m in a seed round joined by AIP Seed.

“ AIP Seed’s portfolio is becoming increasingly international. The Polish founders we invest in build global ventures from the outset, win customers abroad and often develop their companies in the world’s leading technology ecosystems,” says Dariusz Żuk, founder and managing partner of AIP Seed.

The round was led by Crane Venture Partners. Other investors included PXN Ventures, as well as angel investors involved in developing AI infrastructure and companies such as DeepMind.

From Edinburgh to San Francisco

San Francisco-based Expanse was founded by four engineers, all graduates of the University of Edinburgh. All had previously worked on computing systems used in areas including finance and supercomputing centers. One of them, Ismaeel Bashir, developed a system at the University of Edinburgh that predicted demand for computing power. Together with Nikodem Bieniek, Eren Mendi and Yafet Melake, he noticed something striking: even the most advanced computing centers still rely heavily on manual estimates of how much capacity individual tasks will require.

According to Dariusz Żuk, Expanse has the potential to increase the effective utilization of existing computing infrastructure by as much as 30 percent, without requiring comparable spending on new hardware.

“We are looking precisely for projects like this, where Poles with strong technology backgrounds, such as Nikodem Bieniek, co-found companies with global ambitions. This was a highly competitive funding round,” says Dariusz Żuk.

The problem is unused computing power

Expanse is developing software for AI infrastructure designed to help companies make better use of the computing power they already have. The problem is particularly acute with GPUs – processors used for the most demanding artificial-intelligence workloads. Although access to them is costly and limited in many organizations, actual utilization often remains well below capacity.

Before running a task, engineers must determine how many GPUs, CPUs and how much memory will be required, as well as how long the computation will take. Reserving too many resources leaves capacity idle and pushes up costs; reserving too little creates a risk of delays or execution problems. Expanse wants to replace these manual estimates with a model that predicts resource requirements in advance.

According to data cited by the company, around 30 percent of cloud spending is wasted because of poor resource allocation. Expanse also points to research by Microsoft Research, which found relatively low GPU utilization across some of the deep-learning workloads analyzed.

Expert's perspective

Can Expanse turn prediction into real savings?

The idea behind Expanse’s product addresses a genuine market problem. GPUs are expensive, yet their utilization can be surprisingly low. Independent data point in the same direction. In a 2026 report based on data from tens of thousands of Kubernetes clusters, Cast AI found average GPU utilization of just 5 percent. Respondents to a Flexera survey, meanwhile, estimated that 29 percent of spending on IaaS and PaaS is still wasted.

Microsoft Research offers another useful data point. In an analysis of 400 real-world deep-learning jobs, researchers found that 46.03 percent of cases of low GPU utilization were linked to data operations. Nearly 85 percent of the identified problems, meanwhile, could be fixed with relatively minor changes to code or scripts.

Expanse’s idea of predicting GPU and RAM requirements, as well as execution time, before a process is launched therefore has a solid economic rationale. I see the greatest potential in organizations using hundreds of GPUs — in HPC, R&D departments, pharmaceuticals, finance or large AI clusters. With smaller infrastructure, by contrast, the potential savings may not be large enough to justify the cost of implementation.

Another advantage is the option to deploy the solution on-premise, which matters from a security perspective. Code and telemetry do not have to leave the customer’s environment, allowing the organization to retain greater control over data flows. An on-premise solution does not, of course, eliminate every risk, but it does reduce dependence on external services and providers.

The biggest question, however, is how accurate Expanse’s forecasts will prove under real-world and rapidly changing workloads. The market is not empty: tools for GPU scheduling and utilization optimization already exist, including NVIDIA Run.

Expanse’s position will therefore depend not simply on its ability to predict resource requirements, but on whether it can demonstrate that those forecasts translate into measurable savings. If it can, the company may carve out a place for itself in the increasingly competitive market for AI-infrastructure optimization.

Dawid Sowiński

business development specialist and Edge AI solutions architect at Consult Red

Expanse wants to predict computing demand

Expanse’s software analyzes a workload – a specific computing task – before it is launched. On that basis, it predicts how many resources will be required, including GPUs, CPUs, memory and computing time. It then selects the appropriate configuration. According to Expanse, this allows companies to make better use of their existing infrastructure. It can also reduce the risk of failed jobs and postpone costly investment in additional hardware.

As Ismaeel Bashir explains, engineers today often have to estimate for themselves how much computing power their code will require. Expanse wants to automate that process and shift a significant share of those decisions to software.

Why make better use of GPUs?

This matters because GPUs, which are used for some of the most demanding artificial-intelligence workloads, are now among the most expensive components of AI infrastructure. Expanding data centers requires not only buying additional hardware, but also providing sufficient power and cooling. Making better use of existing infrastructure can therefore be a faster and cheaper way to increase available computing capacity than building additional resources from scratch.

Investors are betting on efficiency

What persuaded investors to put money on the table?

Scott Sage, co-founder and partner at Crane Venture Partners, argues that success in AI will depend not only on building the largest computing clusters, but also on how efficiently they are used. In his view, Expanse can help companies manage existing resources more effectively, reduce waste and extend the period for which current infrastructure remains sufficient. In this way, the company is seeking to carve out a new segment of the AI-infrastructure market.

Andy Barrow, managing director at PXN Ventures, takes a similar view. He points out that as AI becomes more widespread, the infrastructure required to support it is becoming increasingly important. One of the biggest challenges is making ever more efficient use of available computing power. In his view, Expanse’s solutions could help companies reduce costs and scale their operations without increasing hardware spending at the same pace.

AIP wants to become increasingly global

The investment in Expanse is intended to support the company both in further product development and in scaling the business. At the same time, it fits AIP Seed’s ambition to build an increasingly international portfolio. That does not, however, mean abandoning the fund’s core principle of investing primarily in Polish founders.

“Our strategy remains unchanged: we invest in Polish founders, and they remain our first choice. We also see many interesting projects outside Poland, but for now this does not mean changing our geographic approach to investing. We believe that the coming years will bring a major wave of excellent global companies founded by Poles. We want to be ready to support them from the earliest stage,” says Dariusz Żuk.

Interactive chart

Pełny ekran

The fund also recently announced an expansion of its supervisory board to include a group of well-known private investors and technology entrepreneurs. They include Marcin Żukowski, co-founder of US company Snowflake, Sebastian Kulczyk, president of Kulczyk Investments, Wiktor Schmidt, associated with Netguru, and Maksymilian Braniecki, former president of Lidl Polska.

As XYZ previously reported, AIP Seed wants to build an evergreen fund, meaning that capital recovered from exits is returned to the fund and can be deployed in further investments. Over the longer term, its managers are also planning to list AIP Seed on the Warsaw Stock Exchange.

“Discussions with the new supervisory-board members have not changed our strategy, but they have raised our level of ambition. They are encouraging us to think even more globally, support founders in building larger companies and create a portfolio that can compete with Europe’s best funds,” Dariusz Żuk adds.

An investor’s view (not involved in the transaction)

Polish funds are becoming bolder abroad

Three factors are driving the growing number of investments by Polish funds in foreign companies, including those with a “Polish connection”, such as a Polish co-founder.

First, there are more and more Poles whose education and professional careers are tied to other countries, making it natural for them to establish companies in Western Europe or the US. Many founders also decide at the pre-seed or seed stage to incorporate outside Poland. This can make it easier both to raise foreign capital, especially in the US, and to win international customers.

Second, founders’ global ambitions are an important investment signal for funds. A team that incorporates in Delaware, actually wins customers abroad and hires international specialists demonstrates that its global ambitions go beyond declarations.

Third, Poland’s VC market is maturing. Funds are developing their capabilities, building international networks and becoming better prepared to invest abroad. Changes to PFR Ventures programs have also helped by allowing part of the capital to be allocated to foreign investments, including – subject to certain conditions – in the US.

The desire to invest abroad has existed for years. Only now, however, is it increasingly translating into actual deals.

Katarzyna Kowalczyk

Managing Partner, 24Ventures

Key Takeaways

1. In the race to develop AI, the challenge is no longer just to build ever-larger infrastructure, but also to use what already exists more efficiently. Expanse wants to forecast computing demand more accurately and reduce the amount of costly capacity that sits idle. The aim is to help companies expand their computing capabilities without increasing spending on additional GPUs, power and cooling at the same pace.
2. Expanse operates in a growing segment of AI infrastructure focused not on adding more resources, but on making better use of those already available. The startup has raised USD 5.3m from investors who believe that competitive advantage can come not only from the scale of infrastructure, but also from how efficiently it is used. The key test, however, will be whether Expanse’s technology can deliver measurable savings for customers.
3. For AIP Seed, the investment in Expanse reflects a strategy of backing Polish founders who are building global companies, rather than limiting itself to businesses operating in Poland. Expanse is headquartered in San Francisco and one of its co-founders, Nikodem Bieniek, is Polish. The fund says it wants to participate in the next wave of global companies founded by Poles and support them from the earliest stages of development.