Building VC-ready AI companies: sustainability as an advantage

This blog post is based on a panel discussion about AI sustainability and investment trends, featuring insights from industry leaders at an AI conference. We utilized AI tools for transcription and to enhance the structure and clarity of the content.

The AI investment is increasingly growing. While major tech companies plan to spend over $300 billion on AI infrastructure in 2025, investors are no longer just asking about powerful models or rapid scalability.  In a recent panel, leaders from the investing, climate, and infrastructure sectors cut through the hype to discuss what “sustainable AI” really means for founders.

The investment reality check

The host began with a simple fact: in 2025, major tech firms, including Amazon, Alphabet, Microsoft, and Meta, are pouring unprecedented resources into AI infrastructure. President Macron recently announced a 109 billion euro investment. But as Resa, General Partner at Partech Global, explains, "What we don't want is for people to keep spending $2 to make a dollar or to hit a wall that will limit their growth."

This perspective reflects a broader understanding that sustainability encompasses both environmental responsibility and business viability. The days of unlimited compute spending are numbered, not just because of environmental concerns, but also because the economics simply don't work in the long term.

The sustainability paradox in high-growth startups

Jean-Baptiste Rudelle, president of the Galion project and former founder of Criteo, shared a crucial insight about the apparent contradiction between hyper-growth and sustainability. "The whole point of a startup is to go in hyper growth, and how can you sustain hyper growth and at the same time be sustainable? There's some kind of mental tension between those two concepts."

His experience at Criteo, where revenues doubled every year for seven consecutive years, provided a practical framework: When you achieve 2x growth in traffic and revenue, you should aim for only 1.5x growth in infrastructure costs. This ratio ensures that your scaling curve creates leverage rather than consuming it.

The cost of ignoring this principle became clear when Criteo hit a ceiling at around $100 million in revenue. The company had to dedicate 18 months to rebuilding its platform from scratch, with the entire technical team focused solely on sustainability improvements while no new products were shipped. As Rudelle noted, "This was the cost to pay for long-term sustainability."

Breaking the misconceptions

Anise, sustainability manager at Reva, highlighted a critical misconception that has shaped the tech industry for years: "For a long time, we thought that tech companies and IT in general weren't part of the problem. We used to say it's the industrial companies that are the problem."

Recent data have challenged this view. A 2019 study by the Shift project revealed that IT represents 4% of global emissions, nearly equal to aviation emissions. Without action, this could grow to 10-11% of global emissions. Even companies like Microsoft, which announced ambitious Net Zero pathways, have seen their emissions grow by 50% due to AI activities.

However, the story isn't entirely negative. AI startups can also be part of the solution, with companies working in healthcare and using AI to optimize energy grids, showing how technology can address social and environmental problems.

Practical implementation

Guillaume, Vice President of Advocacy at Upsun, shared concrete examples of how sustainability translates to business optimization. When their platform reached around $40 million in annual revenue, a dedicated FinOps team reduced cloud bills by 40% in just six months by removing overlooked services and optimizing resource allocation.

The key insight is that sustainability isn't just about reducing carbon footprint—it's about optimizing for cost and ensuring that profit margins are sustainable. This includes tracking hidden expenses, such as bandwidth between regions and backups, which can represent 20-25% of cloud bills.

The tools and metrics that matter

For early-stage companies, the first step isn't complex carbon accounting; it's understanding that sustainability isn't a burden but a competitive advantage. Anise recommends starting with carbon footprint assessments to know where emissions originate, enabling targeted optimization.

For AI-specific applications, the focus shifts to model efficiency. Rudelle emphasized the threshold nature of AI: "If you are below a certain level of quality, the whole thing is useless. If you are below 80% predictability, you can throw the whole thing out of the window."

This creates pressure to increase parameters and computational power, making optimization techniques crucial for effective solutions. Model compression can achieve dramatic results, sometimes reducing model size by a factor of five with only a 1% loss in accuracy. The ability to predict traffic peaks also matters significantly, as dimensioning for maximum load can result in doubling system requirements for minimal gain.

The investment perspective: Beyond the pitch deck

When asked about integrating sustainability into pitch presentations, Resa offered a different perspective: "It's not really about the pitch deck. I want to turn the question around because a pitch deck is just a snapshot of a story they want to tell at a given time."

The more important matter is incorporating the understanding that throwing money at problems isn't sustainable. Investors want to see founders thinking about optimization from the beginning, planning for "what's next" beyond just market size and growth projections.

The example of Facebook's early data center challenges illustrates this point. When California faced a shortage of electricity and water for new data centers, Facebook initiated the Open Compute Project, which developed more efficient server designs and data centers that utilized natural ventilation rather than air conditioning. This wasn't driven by environmental concern but by practical necessity.

Emerging trends and future opportunities

Several trends are reshaping how AI companies approach sustainability:

• Federated Learning: This approach distributes model training across user devices rather than centralizing it in the cloud. This reduces resource consumption while addressing privacy concerns, as data doesn't need to be centralized for training.
• Smart Infrastructure: Companies are leveraging AI to optimize electricity usage in response to grid conditions. In France, for example, carbon intensity varies significantly between 7 PM and 1 AM or between seasons. Innovative dispatch systems can redistribute workloads to regions with cleaner electricity sources without compromising user experience.
• Alternative Hardware: The current reliance on NVIDIA GPUs isn't sustainable in the long term. New architectures, including TPUs and NPU, are emerging, potentially offering better efficiency and reducing the environmental impact of hardware production.

The economic reality

Despite the compelling sustainability narrative, Jean-Baptiste Rudelle provided a sobering economic perspective: "To be completely honest, in the short term, the hard answer is there is very little commercial edge by trying to be low carbon for a startup."

The primary incentive today remains largely reputational, with brand power influencing hiring talent and attracting clients. However, this could change significantly with the introduction of meaningful carbon pricing. “What would really address this in the long term would be a carbon tax. If there is a real serious carbon tax, then businesses will have a real incentive to decarbonize.”

The productivity argument

The panel discussion touched on a crucial economic argument: AI's role in funding the green transition through productivity gains. While productivity improvements from new technologies often take years to materialize in economic statistics, the potential impact on coding and software development could be transformative.

As Rudelle explained, "The area where the productivity is going to be the most spectacular in terms of gains in the short term is going to be in terms of writing code." If the cost of writing code approaches zero, companies can develop customized software tailored to their specific needs, rather than relying on generic SaaS products, which could potentially result in significant productivity gains.

Regulatory and market pressures

The regulatory landscape is pushing sustainability requirements downstream. According to Anise, 87% of companies at the Series C stage have performed carbon footprint assessments. "From Series B onwards, investors make it compulsory to have a climate strategy," she noted, driven partly by regulations like SFDR (Sustainable Finance Disclosure Regulation).

This isn't just about regulatory compliance; it's about risk management. As one panelist noted, "Venture capital is the other name for risk capital, and sustainability is a risk management framework."

Key takeaways for AI startups

The panel's closing recommendations provide a clear roadmap:

1. Act now: It's always less costly to implement sustainable practices early than to repair problems later. Studies by actuaries suggest that the effects of climate change could decrease global GDP by 50% by 2070-2090.
2. Make AI part of the solution: Rather than viewing AI as purely a resource consumption problem, focus on how it can drive efficiency and solve environmental challenges.
3. Think outside the box: Don't rely on traditional scaling approaches. The shift to AI requires rethinking how applications are built and solutions are deployed, much like how cloud technologies have forced architectural changes.
4. Leverage expertise: Don't try to solve infrastructure optimization on your own. Collaborate with specialists and tap into the growing ecosystem of sustainability-focused startups and tools.

Looking forward

The AI sustainability discussion represents more than environmental responsibility; it's about building businesses that can scale efficiently and survive long-term market pressures. As infrastructure costs continue to rise and regulatory requirements increase, companies that integrate sustainability thinking early will have a significant competitive advantage.

The message from investors is clear: demonstrate not just what you're building and how fast you can grow, but how you plan to sustain that growth efficiently. In a world where throwing unlimited resources at problems is no longer viable, optimization becomes the key differentiator.

For AI startups, sustainability isn't a constraint; it's an opportunity to build more resilient, efficient, and ultimately successful businesses. The companies that recognize this shift early will be the ones that attract the right backing and create lasting value in the evolving AI landscape.