Models-as-a-service companies are quickly becoming some of VCs' most sought-after startups. We’re not talking about Deliveroo for Cara Delevigne and Magic Mike; we mean companies like Mistral, Stability AI and Aleph Alpha in Europe, and OpenAI and Cohere in the US, which are selling generative AI to enterprises to help them get their work done more efficiently.
That can be in the form of asking a chatbot the kinds of things you’d normally ask HR, creating onboarding training programmes or surfacing valuable information from large troves of internal documents.
The idea of applying the power of large language models (LLMs) to the modern workplace has been around for a few years but, ever since OpenAI lit the sector alight last year the promise of creating “ChatGPT for your business” has a lot of investors excited.
This year European models-as-a-service companies have raised from big US investors like Sequoia, Lightspeed and New Enterprise Associates.
But, while all of these companies are enjoying the ChatGPT-fuelled explosion of interest in GenAI, they’re taking quite different approaches to building a business. Some are going through the costly process of developing their own models, while others are leaning on existing technology from the likes of OpenAI. There’s also a big range when it comes to the number of clients each has signed up.
We’ve mapped out the most noteworthy models-as-a-service startups based in the continent to compare how their businesses are shaping up.
Here are the models-as-a-service companies to have on your radar.
Mistral
Year founded: 2023
Total investment: €105m
Investors: Lightspeed, Index Ventures, Redpoint, firstminute capital, Sofina Ventures, Headline, Bpifrance, LocalGlobe, Motier Ventures, JCDecaux Holding, Xavier Niel, Eric Schmidt, Rodolphe Saadé
Confirmed clients: Healthtech scaleup Alan (with “more to come”)
Number of employees: 14
Mistral is without doubt some of the hottest property in European AI. Fresh off a monster €105m seed round led by Lightspeed, the company is now setting to work building its own LLMs as a European alternative to the likes of OpenAI’s GPT-4, aimed at European enterprises. It says it will use an open source strategy and business model.
The founders, Timothée Lacroix, Guillaume Lample and Arthur Mensch, spent nearly two decades between them working in AI and machine learning at Meta and Deepmind — a fact the trio were keen to emphasise in their pitch deck, a copy of which was leaked to Sifted.
Sifted take: Mistral is the company that people are talking about in European GenAI, and Sifted understands that the startup is able to pay above market rate for ML and AI engineers after its fundraise, so it will have no issues in attracting the best talent. The team is currently hiring.
Some AI watchers are asking whether giving away equity for VC cash to be spent on training LLMs is a smart move, as open source models continue to improve in quality. But if any startup is going to build a European alternative to the US companies training LLMs, Mistral seems well-placed to be the one to do it.
Stability AI
Year founded: 2019
Total investment: $101m
Investors: Coatue, Lightspeed, O’Shaughnessy Ventures and Sound Ventures
Confirmed clients: "While we cannot name all of our clients publicly, they include Tome and Jasper AI"
Number of employees: 185
Stability AI is, like Mistral, building its own LLM, which is called Stable LM and was released earlier this year. The company also funded the training of Stable Diffusion, one of the best-known text-to-image models that serves as an open source alternative to OpenAI’s Dall-E and US-based Midjourney. Stability burst onto the scene with a $101m investment round led by Lightspeed and Coatue in October 2022 at a $1bn valuation.
Stability is known for having an unusual company structure, with employees distributed all over the world and given a high level of freedom to work on ambitious research. The company’s founder and CEO, Emad Mostaque, has become a vocal proponent of open source GenAI and says that companies using LLMs will be unlikely to want to use “black box” systems owned by big tech companies.
Sifted take: Stability has been dogged by controversy in recent months, with Forbes reporting that the company failed to pay staff and taxes, while a leaked pitch deck seen by Sifted suggested that it may have overstated its capabilities while raising money (Stability denied all claims).
Bloomberg also reports that the company is struggling to raise a new round of funding at an improved valuation, as people begin to question when Stability’s ambitious open-source vision will translate into a solid revenue stream. That said, the company boasts strong talent and compute resources and could still surprise its detractors. Stability will likely need to sign up some clients soon if it’s going to convince investors to recapitalise it.
Sana
Year founded: 2016
Total investment: $85m
Investors: EQT, Menlo, NEA and Workday
Confirmed clients: healthtech Kry, buy now, pay later company Klarna, appliance manufacturer Electrolux, solar infrastructure company Svea Solar, multinational science and tech company Merck
Number of employees: 55
Stockholm-based Sana has an impressive roster of investors and clients on its books, and has been building its suite of AI-powered enterprise products since long before the GenAI boom began. Unlike Mistral and Stability, the startup makes use of third party models like GPT-4, rather than putting money into foundational AI research.
CEO and founder Joel Hellermark, who founded Sana when he was just 19, told Sifted that the company’s recent top-up to its Series B was the result of a “proactive offer” from New Enterprise Associates, one of the world’s largest VCs, which rarely invests in Europe.
Sifted take: Like all the companies in this article, Sana has the task of going up against the likes of Microsoft and Google when it comes to building AI products for the workplace, meaning it will have to be agile as it caters to clients’ needs.
But the startup has already won business with leading scaleups like Kry and Klarna, giving it a head-start against some of its rivals. Winning established industrial clients like Electrolux and Merck also shows that Sana has built a product that’s creating value outside of the tech community — a useful asset amid a funding downturn.
Aleph Alpha
Year founded: 2019
Total investment: €28.3m
Investors: Earlybird VC, Lakestar, Unternehmertum Venture Capital Partners, 468 Capital, LEA Ventures, Cavalry Ventures
Confirmed clients: BWI (the IT provider to the German military), the City of Heidelberg, State Ministry of Baden-Württemberg
Number of employees: 60
Aleph Alpha, based in Heidelberg, is another GenAI company that’s been building since long before the hype began. The company has built its own model called Luminous and has put a lot of emphasis on combating hallucination and researching ways to make its systems more reliable.
Founder and CEO Jonas Andrulis tells Sifted that Aleph Alpha is focusing on what he calls “critical enterprises” — organisations like law firms, healthcare providers and banks, which rely heavily on trustable, accurate information. The company offers LLM services in five languages: English, German, French, Italian and Spanish.
Sifted take: Aleph Alpha has had a bit of a head-start in signing up clients compared with competitors building LLMs from scratch, and is already working with some established institutions. This first-mover advantage, and its focus on multi-lingual services, will help the company as it comes under more pressure from rivals like Mistral touting a similar “European models for European organisations” message.
Dust
Year founded: 2023
Total investment: $5.5m
Investors: Sequoia, Seedcamp, Connect Ventures, Remote First Capital, Motier Ventures
Confirmed clients: Six “design partners” (not yet paying customers)
Number of employees: 8
Cofounded by former OpenAI research engineer Stanislas Polu, Dust was the talk of VC town ahead of raising a $5m seed round led by Sequoia in June. Like Sana, the company is using a range of third-party LLMs, with Polu telling Sifted that there’s still a lot of unrealised value to be extracted from models like GPT-4.
Dust is now working with six “design partners” — other organisations that all have more than 70 employees — to find new use cases for LLMs in the workplace. Polu says that we’re still in the very early stages of working out how GenAI can be applied to businesses and there’s still a lot of work to be done at the product layer.
Sifted take: Dust raising a more typical seed round than its Parisian neighbour Mistral demonstrates one of the appeals of building a GenAI company focusing on product, rather than foundational research: it’s much cheaper to do.
Polu’s cofounder Gabriel Hubert previously developed products at payment provider Stripe and healthtech Alan, giving the founding team a strong mix of experience. Dust tells Sifted that it’s getting a lot of inbound requests from companies that want to use its services, and that there’s a clear route for its current design partners to become paying customers.
Silo AI
Year founded: 2017
Total investment: Not disclosed
Investors: Altor Equity Partners
Confirmed clients: Allianz, Sandvik, Tietoevry, Happeo and “10-15 other clients”
Number of employees: 300
SiloAI, unlike many of the companies on this list, is profitable. It took private equity investment from Altor Equity Partners in 2022, but until then had largely been bootstrapped.
In April the company launched a new GenAI offering to businesses based on LLMs, building on the services it’s built using other types of AI models over a number of years. The company uses a mix of third-party and proprietary models.
Sifted take: Silo, which has been creating AI products for clients since 2017, has a big advantage in its work with big customers like Allianz and Sandvik, which gives it a stamp of legitimacy and trust that others will need to build. It can also sell its LLM-based services to pre-existing customers. Silo tells Sifted it’s already pulled in €23m in revenue in 2023.
Lighton
Year founded: 2016
Total investment: €5m (including grants)
Investors: Quantonation, Anorak Ventures, Otium Capital
Confirmed clients: 10
Number of employees: 20
Lighton is another profitable GenAI company that’s been offering models to enterprise clients for a number of years. The startup began life developing hardware components for high-performance computers, before pivoting into building LLMs in July 2020.
Lighton researchers contributed to the development of Bloom — one of the most popular open source LLMs — and, as far as Sifted knows, is the first European company to have built a GenAI model in Arabic, for a client in the UAE.
Sifted take: Lighton’s revenue from its hardware business — in combination with grant funding — has allowed the company to keep its fundraising relatively lean, meaning it’s flown somewhat under the radar as rivals' megarounds poured in.
But, like the other startups in this list that have been running for a few years, Lighton has already accumulated a strong list of clients. The question will be to what extent it can defend its position, as newer, better-funded competitors establish themselves in the market.
Nyonic
Year founded: 2023
Total investment: NA
Investors: NA
Confirmed clients: Company did not respond
Number of employees: Six (according to Dealroom)
Berlin-based Nyonic is one of the newest kids on the block in European GenAI and is developing and training its own LLM. The company will sell its services to industrial clients in Europe and focus on providing services in multiple European languages.
Nyonic is led by an impressive team that includes Johannes Otterbach, who worked on OpenAI’s technical team for two years, and Vanessa Cann, managing director of the German AI Association. Cann tells Sifted that the company is working with industry associations to get its hands on high-quality, industry-specific data.
Sifted take: It’s early days for Nyonic and the company has its work cut out as it develops its own LLMs in a race to compete with the other names on this list. But the company has a strong mix of commercial and technical talent in its funding team, and a clear strategy for differentiating itself by focusing on industrial clients. Given Europe’s strength in industries like manufacturing, it could be a smart bet.
