Latest Interviews

Balderton, Altimeter, TVG, Coatue, Alpha & Harmonic: Investment in AI Who Gets Funded and Why

Max, Avid, Antoine, Apoorv, Maliki, Bernard

A recent analysis reveals that generative AI revenue is currently dominated by semiconductor giants like NVIDIA, creating an inverted economic pyramid where early startups face massive infrastructure costs and power grid bottlenecks. Venture capital strategies are shifting toward large, high-conviction checks focused on deep infrastructure and energy assets, while successful founders must combine deep domain expertise with technical literacy to navigate complex enterprise adoption. Despite these challenges, high-velocity "AI-first" companies like Higgsfield and Lovable are achieving rapid scale and superior margins by leveraging internal AI efficiency before productizing their services.

A Frieside chat with Rafay & BUZZ HPC: Sovereign AI in Action From Infrastructure to Impact

Rafay, BUZZ HPC, Haseeb, Craig

BuzzHPC positions itself as a sovereign AI provider for Canadian enterprises by operating an integrated stack across Sweden and Canada that delivers over 135,000 GPUs while ensuring strict data residency and compliance with regulations like the U.S. Cloud Act. By transitioning from declining bare metal margins to value-added model-as-a-service offerings, the platform achieves revenue per H100 of $3.50 to $5.00 through self-service consumption, multi-tenancy, and enterprise-grade SLAs. This approach satisfies modern developer expectations by offering granular isolation from single units to 1,000+ GPU clusters, enabling efficient end-to-end project hosting without the friction of data movement between hyperscalers and domestic clouds.

Base Power Raises $1B led by Addition, with Trust Ventures, Lightspeed, a16z, Altimeter, Elad Gil

Elad Gil, Zach Dell, Justin Lopas, Molly O'Shea

Base secured an oversubscribed $1 billion Series C led by Addition, accelerating from its previous $200 million round to fund its "Speed to Power" product designed to bypass traditional grid interconnection delays for data centers. Backed by strategic investors like Lennar and Starwood alongside founders from SpaceX and Tesla, the company is deploying gigawatt-scale battery capacity from its Austin "Factory One" at a target rate of 100 megawatts per month. This capital infusion supports a strategic pivot to replace fossil fuels with cost-efficient solar and storage, aiming to solve critical energy bottlenecks for AI development while expanding operations into high-solar states like California and Florida.

A fireside Chat with Mati Staniszewski, CEO & Co-Founder of Eleven Labs ft. Balderton Capital

Mati Staniszewski, James Wise

Founded by Mati Kovacs and Peter, Eleven Labs has evolved from addressing Polish film dubbing limitations into a global AI voice leader operating with offices in the US, Europe, India, and Japan. The company currently executes a dual strategy of frontier research and application development, delivering production-ready voice agents for healthcare scheduling and interactive entertainment while distinguishing itself through high-fidelity dialect recognition and a creator marketplace that has paid over $5 million. Looking forward, leadership is prioritizing a three-tier trust framework for safety and a technical pivot toward combining cascaded reliability with end-to-end Duplex models to achieve human-like latency in the coming year.

Scaleway & SemiAnalysis: Unlocking AI's Potential Insights from Deploying Europe's Most Advanced

Scaleway launched Europe's largest Hopper GPU cluster in 2023 to secure data sovereignty and prevent U.S. hyperscaler dominance, now hosting major AI players like Mistral for training and inference. The provider is advancing "Project ION," a consortium-backed initiative aiming to build capacity equivalent to 280,000 H100 chips with private capital while seeking government demand assurance rather than direct subsidies. Operating across five European regions, Scaleway positions itself as a full-stack public cloud solution that prioritizes legal immunity from U.S. laws over domestic hardware manufacturing, targeting a consolidated market of EU-based public and private AI consumers.

Arm, PsiQuantum, Intel, NextSilicon, Lightspeed: Moore’s Law vs AI’s Appetite The Compute Collision

Jean-Anne Philippe, Will Abbey, Elad, Pete Chabot

Intel's Jean-Luc Philippe, ARM's Will Abbey, and experts from Next Silicon and PsiQuantum convened to analyze the shift from traditional transistor scaling toward software-hardware co-design and heterogeneous computing architectures. The discussion highlighted critical advancements in quantum error correction and RISC-V integration while emphasizing that future AI infrastructure requires billions in capital to support hybrid edge-cloud systems and post-quantum cryptography. Industry leaders concluded that the end of Moore's Law necessitates fundamental architectural innovation over brute-force scaling, with a consensus that value is forged through massive, power-intensive supercomputing deployments rather than isolated component optimization.

Poolside AI, IREN, Forbes: From 10x to 100x Building AI Systems for Real World Scale

Emma Allen, Dennis Skriniop, Isa Kent

Dennis Skriniop of Iron and Isa Kent of Poolside discuss the critical infrastructure and energy requirements for scaling foundation models toward human-level intelligence. Kent outlines Poolside's $500 million investment in automating a "model factory" that executes 1,800 monthly experiments, emphasizing that 80% of success relies on distributed systems engineering rather than algorithmic discovery. Skriniop and Kent further analyze the necessity of hybrid power grids, disaggregated storage architectures, and the emerging role of specialized "Neo-Clouds" to address the exponential complexity of operating clusters exceeding 10,000 GPUs.

Lin Qiao, Co-Founder & CEO, Fireworks AI: Open Models: The future of AI agent development

Lin Qiao

Fireworks CEO Lin positions 2025 as a pivotal year for AI agents, highlighting the platform's deployment for major enterprises like Cursor, Uber, and UiPath through a unique "data flying wheel" strategy that aligns model distributions with proprietary product data. The service scales 5 trillion tokens daily across 1,000 models while utilizing multilora compression and reinforcement fine-tuning to achieve up to 15x performance gains in coding and research tasks. By offering enterprise-grade security and bringing custom models to production with 100x faster iteration speeds, Fireworks enables organizations to transform AI from a utility into a core intellectual property asset.

WEKA, Intel, Graphite, InstaDeep & AWS: Scaling Without Crashing. AI Framework for Enterprise Growth

Eduardo, Mary, Damien, Val, Karim, Murray, Fula Ousmane

The event details the industry shift from experimental AI in 2023 to the scaling of autonomous agents in 2024-2025, driven by urgent economic pressures that are forcing a 75% reduction in cached input pricing. Industry leaders outline a strategic framework prioritizing top-level governance and data transparency to mitigate security risks while navigating a market where compute cost competitiveness will determine the survival of AGI-era businesses. Specific sector applications, ranging from diagnostic health agents to AI-driven code review systems, illustrate how organizations must integrate these tools as central transformation enablers rather than isolated initiatives.

Inception Fund & AI: From Inception to Exit Building AI Giants from Europe and the Nordics

Peter Sarlin, Oliver

Former professor Peter Sarlin founded the private AI lab Silo AI in 2017 and led the bootstrapped company to a $700 million exit when acquired by AMD in 2022, establishing the largest pure-play AI acquisition in Europe. Sarlin attributes this success to a contrarian philosophy that prioritized functional machine learning over hype while hiring primarily from academia, resulting in a workforce of over 220 AI scientists before the strategic sale. Beyond this specific case study, Sarlin highlights that the fragmentation of the European single market remains the primary barrier preventing regional startups from scaling into trillion-dollar enterprises, forcing founders to adopt a global-first strategy despite Europe's strengths in industrial AI applications.

Alex Ratner, CEO & Co-Founder, Snorkel AI: Data Centric AI in the Agentic Era

Alex Ratner

Alex, co-founder of Snorkel.ai, argues that specialized agentic systems require robust evaluation benchmarks composed of prompt datasets, rubrics, and verifiers to serve as functional product requirements. His company provides data operations platforms to major US banks and federal agencies while demonstrating that custom-trained models using programmatic evaluators can outperform GPT-4 on complex insurance underwriting tasks. This approach establishes high-quality data and verification logic as the foundational prerequisites for optimizing Reinforcement Learning in enterprise environments.

OVH Cloud, ClickHouse, Neon, Aiven, Airbyte & Commit: Open Software Infrastructure 2.0

Abel, German, Oskari, Heike, Michel, Alexei, Francois Beaufort

Executives from Aiven, Neon, Airbyte, ClickHouse, and OVHcloud converged to discuss how open-source data platforms are securing $10 billion in combined valuations while addressing enterprise needs for data sovereignty and interoperability. The panel detailed how the AI boom is driving a shift from human-led development to agentic provisioning, creating demand for storage-centric architectures that support short-lived, unstructured data workloads. Looking ahead, industry leaders predict a five-year trajectory where hundreds of millions of citizen developers leverage commoditized AI to compress development cycles, necessitating hybrid infrastructure strategies that balance local model execution with scalable cloud data management.

Alix, SAP, Daytona, Feedly, Prosus & Handelsblatt: The Future Agentic AI & Autonomous Workflows

Alix, Edwin Kodabachan, Paul Van der Boor, Alexander Maisor, Emmanuel Casimates, Ivan Bourazin

Experts at a recent industry gathering defined Agentic AI as a shift from simple request-response models to autonomous systems capable of multi-step planning, emphasizing that security and governance remain critical barriers requiring strict human oversight. SAP and other leaders proposed frameworks like "agentic quality controls" and locally hosted environments to manage agent identities and comply with regulations, while anticipating a workforce evolution where humans supervise "digital teammates" rather than performing manual tasks. Despite concerns over the hype cycle, participants agreed that organizations with mature data and processes will soon leverage these tools to exponentially increase productivity, provided they implement robust guardrails to direct this technological capability toward tangible business value.

Oege De Moor (XBOW) & Apoorv Agrawal (Altimeter): Hackers with GPUs Offensive Security in the AI Era

Oege De Moor, Apoorv Agrawal, Uhey Dhimur

Altimeter Capital-backed startup Expo has deployed an autonomous AI agent that ranks as the world's top ethical hacker, identifying over 1,200 vulnerabilities including critical zero-days in enterprise giants like Palo Alto Networks and Salesforce within a single week. This superhuman capability forces a strategic industry shift toward AI-driven defense to combat a "skill shortage" in malicious actors that has enabled nation-states to automate attacks with unprecedented speed. While the market anticipates a major cyber incident within twelve months due to rising AI-generated code risks, human security roles are expected to evolve from technical execution to the strategic orchestration of these autonomous systems.

Fractile AI, Rogo, Scribble VC, San Francisco Compute: The AI Trillion-Dollar Opportunity

Eric Manisse, Elizabeth Weill, Walter Goodwin, John Willett

San Francisco Compute CEO Eric Manisse, Scribble Ventures' Elizabeth Weill, and founders Walter Goodwin and John Willett convened to analyze strategies for scaling AI infrastructure, hardware, and enterprise solutions amidst rapid technological evolution. The panel detailed critical operational frameworks, including balancing hardware development timelines with software agility, mitigating hallucination risks in financial applications, and proactively shaping export control regulations rather than merely complying with them. Experts concluded that long-term success in this sector relies on specialized human expertise for deep tech, the creation of durable data moats, and a shift toward treating global compute as a utility to drive the next industrial revolution.