AI startup gold rush: Inside the high-stakes world of AI investments

Nvidia might have become the world’s most valuable company for a brief period this year – but when it comes to investments in AI, 2024 represented yet another landmark 12 months for venture capital flowing into AI startups around the globe.

The likes of Anthropic, Scale AI, Cohere, Mistral, Figure and Elon Musk’s xAI all raised hundreds of millions of dollars in 2024 to fund their AI development and expansion efforts.

However, nothing quite compared to OpenAI’s October funding raise of $6.6 billion in what was the largest venture capital round in history. Securing the firm a valuation of $157 billion, that’s more than the current market caps of big-name companies like IBM, Verizon and Qualcomm.

The cash being injected into AI startups is at ludicrous levels, increasing exponentially following the mainstream market’s interest in the technology in the wake of launches of generative AI applications like ChatGPT and Stable Diffusion.

Is this level of investment sustainable, or could we see the bubble burst with emerging spaces like quantum computing turning heads?

Capacity aims to find out if AI will remain the next big thing to invest in going into 2025, and whether this trend applies beyond Silicon Valley.

There’s a myriad of venture capital firms investing in AI startups. One is US-based Costanoa Ventures, which backs startups in early stages like seed and series A rounds.

Martina Lauchengco, a partner at Costanoa, explains that the firm will only invest in startups that are building on top of AI in some way as their foundational application architecture — not those solely wrapping an application around an OpenAI model.

Results published by the venture capital firm from a survey of startup founders showed that while nine out of 10 (89%) are using some version of GPT in their startups, more than half said they were using four or more foundational models.

Lauchengco says the results were akin to a “litmus test” to uncover the more sophisticated startup propositions. Rather than looking for business-minded founders who want large language models to perform entire tasks, Lauchengco explains that the team at Costanoa is looking at “the next layer below”.

Startups that have passed Costanoa’s litmus test include Beacon AI, which is attempting to create an “R2-D2 for pilots” to aid flight safety, and Assured, a claims intelligence platform that helps ingest and augment structured claims data to enhance decision-making.

But the firm is not investing solely in the US, with a notable member of Costanoa’s portfolio being Rerun, a Swedish startup creating software development kits for visualising multimodal data for applications like computer vision and robotics.

Spotting such startups isn’t an easy task, and it helps if you have experts who understand the technology to avoid investors throwing money at the wrong pitches.

Among the ranks at Costanoa are technologists, including an ex-product manager at data and AI firm Databricks. Having actual data scientists and people who understand the technology helps the investors pumping money into all these AI startups to know their cash is going to the right places.

“I think you need someone that has that amount of specialisation to be able to understand and help separate the wheat from the chaff,” Lauchengco says. “We have gotten as far as an individual partner likes [a startup] and then it gets pitched to the broader team, where we have all these specialists that are kicking the tyres on the technical architecture.

“This collaborative approach helps us make sure that we are making sound investment decisions.”

Much of the focus has been on AI startups emerging from Silicon Valley and other Western innovation hubs.

But in 2024, it's not only Western firms that are amassing significant funding. Beijing-based Moonshot AI has raised over $1 billion from investors such as Alibaba, while Baichuan, founded by former Tencent engineer Wang Xiaochuan, secured $688 million in a series A funding round. Additionally, MiniMax, which is developing a Chinese ChatGPT-like solution, obtained $600 million in series B funding, also with support from Alibaba.

While the story might seem similar in Asia, there are some subtle and not-so-subtle differences.

Lian Jye Su, chief AI analyst at Omdia, explains that for markets like China, information on AI investments and funding has always been discreet and somewhat secretive, with startups often emerging from large tech companies as “pet projects” by employees from major firms.

He outlines that Asian generative AI startups are increasingly entering the media and entertainment, image and video generation, and enterprise software markets, with some firms only emerging in the past 12 months.

While some Chinese startups are gaining traction, they face significant regulatory hurdles that limit their commercial reach, chiefly rules by the country’s internet watchdog requiring AI models to be approved before public release.

Beyond regulatory hurdles, Su suggests China’s broader economic challenges are impacting optimism around AI startups.

“The economy is not trending in the right direction this year, and that’s why you have a stimulus plan,” he notes, adding that media and investors alike are “not hyping up these companies as much as they were last year”.

Su also points to a growing sense of caution in China’s AI sector, describing it as a “bubble in the making”. He explains that without a very clear commercialisation strategy, 2025 may mark a turning point as investors become more “realistic about their gen-AI spending”, which could “eventually hurt the startups that don’t quite have a commercial strategy”.

Southeast Asia, in contrast, is emerging as a growth area for cloud and AI services as hyperscalers like Microsoft invest in infrastructure and partnerships across countries such as Indonesia and Malaysia.

Su outlines that companies like Microsoft are looking to capitalise on lower saturation levels in the region to fuel their growth, often collaborating with local operators to establish infrastructure and introduce AI solutions.

Despite the opportunity, there are regulatory hurdles for the wider Southeast Asian market. Su points out that the region’s fragmented economic landscape and varying levels of digital maturity create challenges. Markets such as Singapore and Malaysia, for instance, balance both Western and Chinese technology influences, making them competitive battlegrounds for hyperscalers.

What’s the actual impact on startups, especially those that don’t share the glamour of the likes of OpenAI?

Victor Kristof is the co-founder and CEO of DemoSquare, a Swiss-based startup automating the monitoring and analysis of political and regulatory data with AI.

In August 2024, DemoSquare raised 1.2 million Swiss francs (US$1.35 million) in pre-seed funding, with QBIT Capital leading the round.

Kristof outlines the funding raise process from a startup’s perspective: “We met with the fund manager of QBIT Capital in June 2023 at an investment event in Switzerland, but it was very early for us: we had no customers, and it was just me and my co-founder Jérémie [Rappaz].

“It was nine months of frequent discussions with the fund and their analyst, giving them regular updates. We got more customers and traction through partnerships with media in Switzerland – and after, they decided that they were seeing enough momentum to make a decision.”

After that initial investment, Kristof and the early DemoSquare team went into what he describes as “full fundraising mode”. In one month alone, they spoke to around 50 venture capital firms.

“It was non-stop talking to VCs. They were asking for complementary information, follow-up calls and in-person meetings. It took a lot of time and energy and was much more stressful, despite already having a lead investor.”

Kristof and Rappaz began working together on DemoSquare the same month OpenAI released ChatGPT.

The pair have academic backgrounds and have built machine-learning and predictive models. They weren’t investors like OpenAI CEO Sam Altman, but built the startup from the ground up, applying the technology to the use case first.

While OpenAI has achieved fame and fortune, DemoSquare has also experienced significant growth over the past two years.

The team has expanded from two to eight members and has exciting plans to extend the firm’s reach beyond Switzerland, with a particular focus on Brussels.

Kristof mentions that although much smaller amounts of capital have been raised compared to OpenAI, this funding has been essential for keeping the startup and his team’s dreams alive.

AI spend is only ever going to increase – though plateauing towards investing smartly rather than the scattergun approaches seen a few years ago.

Startups like Beacon, Baichuan and DemoSquare, which are cleverly applying AI to specific use cases and applications, are likely to benefit most from venture firms like Costanoa looking to back cutting-edge AI players.

Kristof sums it up: “We’ve seen a lot of companies just wrapping their data sets around OpenAI and other LLMs. You have to be more innovative and more creative about how you want to include AI in your company and your products.”