Poetiq’s $45.8 million seed round, announced Wednesday, comes with proof that six former Google DeepMind researchers have cracked a fundamental barrier in AI reasoning—achieving 75% accuracy on the notoriously difficult ARC-AGI-2 benchmark, a 16-percentage-point leap beyond previous state-of-the-art.

The achievement matters now because enterprises are drowning in a $30-40 billion AI investment crisis. An MIT study published in August 2025 found that 95% of organizations are “getting zero return” on their GenAI deployments, with most failures traced to LLMs’ inability to handle real-world reasoning tasks that require more than pattern matching.

Limy emerged from stealth today with $10 million in funding to address what its founders call the “agentic web” infrastructure gap—the disconnect between traditional user-centric analytics and the AI agent behaviors that increasingly drive online discovery and commerce.

The funding round, led by Flybridge with participation from a16z speedrun, signals investor recognition of a fundamental shift: As AI agents become the primary interface for information discovery, brands need entirely new infrastructure to understand and optimize for agent behavior rather than human clicks.

Fiddler AI closed a $30M Series C led by RPS Ventures, bringing total funding to $100M as enterprises confront the reality that AI agents require fundamentally different infrastructure than traditional software—not just monitoring, but active control.

The governance gap has real consequences. Agent failures can trigger regulatory fines, legal exposure, and brand damage, driving rapid adoption among Fortune 500 companies where AI oversight isn’t optional. Fiddler achieved 4x revenue growth over 18 months and earned AWS Pattern Partners status as organizations realize that deploying autonomous agents without proper controls creates existential risk.

LiveKit raised $100 million Series C at a $1 billion valuation, led by Index Ventures with Salesforce Ventures, Hanabi Capital, Altimeter, and Redpoint Ventures participating. The open-source-born infrastructure provider powers the real-time voice capabilities behind OpenAI’s ChatGPT voice mode, xAI’s Grok Voice Agent API for Tesla vehicles, and hundreds of enterprise voice AI deployments.

The funding addresses a fundamental infrastructure bottleneck: building production-ready voice AI requires orchestrating speech-to-text, language models, text-to-speech, and real-time communication protocols—a complex technical challenge that most companies struggle to solve in-house. LiveKit abstracts this complexity into a unified platform, turning months of infrastructure development into minutes of deployment.

Inferact launched Wednesday with $150 million in seed funding at an $800 million valuation to commercialize vLLM, the open-source inference engine that reduces AI deployment costs by up to 70%. The round, co-led by Andreessen Horowitz and Lightspeed Venture Partners, represents one of the largest seed valuations ever and signals a fundamental shift in AI industry priorities from model training to deployment optimization.

The infrastructure bottleneck is real. Organizations deploying AI applications are discovering that inference costs—the expense of running trained models to generate outputs—often exceed training expenses over a product’s lifetime. Companies like Stripe report 70% cost reductions using vLLM, while the technology enables significantly faster processing and higher hardware utilization across the AI stack.

Railway raised $100 million in Series B funding as AI coding assistants create a fundamental mismatch between code generation speed and deployment infrastructure, with the company claiming sub-second deployments versus the 2-3 minutes required by traditional cloud tools.

The funding round, led by TQ Ventures with participation from FPV Ventures, Redpoint, and Unusual Ventures, positions Railway to challenge Amazon Web Services, Google Cloud, and Microsoft Azure with infrastructure purpose-built for the AI development era. Railway has attracted 2 million developers with zero marketing spend, processing over 10 million deployments monthly and handling more than one trillion requests through its edge network.

Upscale AI raised $200 million in Series A funding at a $1 billion+ valuation, reaching unicorn status just months after its $100 million seed round. The rapid ascent reflects growing industry consensus that networking has become the critical bottleneck for scaling AI systems.

The Santa Clara startup’s meteoric rise signals a fundamental shift in AI infrastructure priorities. While the industry has focused intensively on compute and storage, networking—the connective tissue that enables AI systems to function as unified clusters—has emerged as the next frontier requiring purpose-built solutions rather than retrofitted legacy approaches.

Elyos AI just closed a $13 million Series A round led by Blackbird Ventures to tackle one of the most significant and overlooked bottlenecks in AI deployment: the operational infrastructure of traditional industries.

The timing is critical. As the AI infrastructure boom creates unprecedented demand for skilled trades workers—from electricians building data centers to HVAC technicians maintaining them—the very businesses supplying this labor are being choked by analog operational friction, preventing them from scaling to meet demand.

Parloa’s $350 million Series D round values the Berlin-based customer service AI startup at $3 billion, tripling its valuation in just eight months from May 2025’s $1 billion mark. The massive funding signals unprecedented capital flowing into customer service infrastructure as enterprises seek to automate the 17 million contact center agents worldwide.

This infrastructure transformation addresses a critical bottleneck: human customer service operations that cannot scale with digital business growth. Enterprise customers including Allianz, Booking.com, HealthEquity, SAP, and Swiss Life are already deploying Parloa’s AI agents to handle customer interactions that previously required human representatives.

Deepgram’s $130 million Series C at a $1.3 billion valuation positions the company as the foundational API platform for the emerging B2B Voice AI economy, with over 1,300 organizations already building voice AI functionality on its real-time infrastructure.

Led by AVP with participation from existing investors including Alkeon, Tiger, Wing, and strategic partners like Twilio, ServiceNow, and SAP, the funding accelerates Deepgram’s mission to become the “Stripe of Voice AI”—delivering the critical infrastructure layer that enables billions of simultaneous voice conversations at human-level naturalness and reliability.