At this year’s Web Summit (May 11-14), Lu Zhang, founder and managing partner of Fusion Fund, said she sees a very different AI conversation taking shape.
“The last time I was at the Vancouver event, the discussion was more about better AI performance, better benchmarks,” she told Investing News Network at the launch of the event. “But it’s very pragmatic about the deployment of AI this year.”
According to Zhang, the change is widespread and obvious. Security, compliance, governance and cost, topics that were once in the background, now dominate the discussion, especially in highly regulated sectors like health care, financial services and insurance.
“All these practical problems become the focus of discussion,” he said.
For Zhang, the most important sign that AI is maturing is that industry players are finally talking about cost.
She presents today’s AI race as a competition over total costs. “When we talk about cost… computation, energy consumption, data consumption, everything together, inference optimization, architecture design – how to make AI solutions better and cheaper,” she adds. “I like this direction… I’m really happy.”
The example he highlighted is Vancouver-based company SenseNet, which has created a dedicated AI solution that integrates multiple data sources, including satellite imagery and proprietary gas sensor data, to detect wildfires before they become visible.
“They’re not only working with the government, they’re also working with utility companies, carriers, insurance businesses, because they also suffer a lot of losses, and they have to pay a lot of fines if they cause wildfires.”
Zhang first met the founder at last year’s event when the company was pre-revenue. Less than a year later, SenseNet has achieved “nearly double-digit million” revenues, she says. For them, SenseNet is a proof point that when AI is tied to a clear, high-stakes problem and when the deployment economics work, development can be extremely fast.
Changes in physical data
Zhang said he has seen a fundamental change in the AI landscape. He commented, “The narrative is changing, shifting from language models to world models to physical AI and from chat to agentic.”
While standardized digital language data may be abundant, the specialized information needed for physical AI, including robotics, sensing, and real-world interactions, remains scarce. “We have about 50 per cent of the infrastructure and we do 50 per cent of the calculations. Data is not even 10 per cent.”
This deficit reflects what she considers one of the major innovation prospects in the current AI field: the rise of firms focused on tactile sensors, novel data platforms, and similar pipelines designed for 3D real-world data.
Major corporations are beginning to appreciate the importance of this data. Zhang reports that among the 45 CTOs in her professional circle spanning the logistics, semiconductor and manufacturing industries, there is a growing demand for startups that can help them harvest and organize their industrial data assets.
On the energy side, Zhang argues that the biggest burden in AI systems is not necessarily computation, but moving data around. “That portion of energy consumption is actually 100 times higher than calculated,” he said.
This becomes particularly acute for three-dimensional physical data, which is much larger than text. In Zhang’s opinion, as physical AI and robotics scale, edge computing will become more essential. Without that change, she doesn’t believe widespread deployment will be sustainable.
Where is the ROI?
The same practical lens applies when Zhang looks at AI in healthcare. While most of the public discussion focuses on AI for drug discovery, she says the real budget, and therefore the real return on investment, sits in optimizing clinical trial outcomes.
His company has already invested in a founder using AI to improve clinical trials and several companies building vertical AI models for specific therapeutic areas.
He cited as an example a company developing vertical AI models for cell therapy, effectively creating a digital twin of a human cell.
“Just think about the different signals – they can potentially mimic the entire process of human cell development under different disease conditions.”
He said the company recently announced a major partnership with a large European pharmaceutical firm focused on Parkinson’s disease, involving assets worth “a few hundred million dollars.”
Zhang also pointed to other portfolio companies, including one that uses microglial cells and vertical AI to treat Parkinson’s and dementia, and another that uses high-density ultrasound with AI to treat depression through a non-invasive, highly targeted approach.
For Zhang, the key development in healthcare is that the value chain is finally becoming complete. Early diagnosis, once given less importance by the health care system, is now increasingly being linked to targeted treatment plans.
He argues that this is why “it’s a great time for AI healthcare.”
Don’t forget to follow us @INN_Technology For real time news updates!
Securities Disclosure: I, Megan Seiter, do not have any direct investment interest in any of the companies mentioned in this article.
