Mega-Series A Rounds Are Redrawing the AI Funding Map
AI startup funding rounds are entering a new phase, with early-stage companies now commanding valuations once reserved for late growth stages. Hark’s oversubscribed Series A of more than USD 700 million (approx. RM3.22 billion) at a USD 6 billion (approx. RM27.6 billion) valuation, alongside Decart’s USD 300 million (approx. RM1.38 billion) raise, reflect investors’ willingness to back ambitious, vertically integrated AI plays. These are not small bets on narrow tools; they are ecosystem-scale wagers on platforms, infrastructure, and hardware that can power the next generation of AI-native products. The high Series A valuation levels signal that investors see the market maturing past experimental pilots and into practical deployment. Capital is concentrating around startups that can deliver not only cutting-edge research but also reliable real-time AI inference, differentiated IP across the stack, and clear paths to monetisation across multiple industries.
Hark’s Universal AI Interface and Hardware-First Vision
Hark is positioning itself as a universal interface between humans and machines, deliberately spanning AI models, software, and AI hardware development rather than focusing on a single layer of the stack. Backed by investors including Nvidia, AMD Ventures, Qualcomm Ventures, Intel Capital, and Salesforce Ventures, the company aims to build agentic, multimodal systems that remember users, retain context across interactions, and integrate with everyday products and services. Founder and CEO Brett Adcock describes the goal as an AI that “actually knows you, speaks your language” and lives on hardware designed specifically for highly personalised use. With around 70 employees and access to a new NVIDIA B200 data center, Hark is preparing to launch its first models later this summer. The funding priorities clearly emphasise massive compute capacity and tightly coupled hardware-software design, underlining a belief that differentiated AI performance will depend on vertically integrated systems.
Decart Bets on Real-Time Video, World Models, and Agentic Inference
Decart is targeting the real-time frontier of AI, focusing on video understanding and world models that can power interactive and physical applications. Its core product, DOS 2.0, delivers over 1,600 tokens per second for agentic inference—around eight times the industry average—alongside full-HD video at up to 100 frames per second across NVIDIA, Google, and Amazon hardware. Built on this stack, the Lucy model line supports real-time immersive experiences in gaming, e-commerce, and advertising with sub-30-millisecond responses, while Oasis is aimed at robotics and autonomous vehicles via physically accurate simulation. Investors such as Radical Ventures, NVIDIA, Adobe Ventures, Toyota Ventures, and eBay Ventures, plus strategic customer Amazon, are effectively funding an infrastructure layer for real-time AI inference. Decart’s priorities highlight that performance, latency, and cross-cloud operability are becoming as critical as raw model size in the next wave of AI deployment.
Beyond Language Models: Why Investors Are Leaning Into Multimodal and Real-Time AI
The latest AI startup funding rounds underscore a shift from text-centric models toward multimodal, interactive systems embedded in real-world workflows. Hark’s focus on agentic, multimodal assistants that manage a user’s digital world, and Decart’s emphasis on real-time video and world models, both point to AI experiences that are continuous, context-rich, and visually grounded. Investors are backing platforms that can move seamlessly from language to images, video, and physical simulation, enabling use cases such as autonomous agents, immersive commerce, and robotics. This evolution demands more than model innovation; it requires robust inference infrastructure, dedicated hardware, and software that can orchestrate complex interactions. In turn, the high Series A valuation levels suggest confidence that monetisation will flow from these practical deployments, as enterprises look for AI systems that can handle latency-sensitive, mission-critical tasks instead of isolated chatbot-style interactions.
What Billion-Dollar Series A Valuations Signal About the AI Market’s Next Phase
Series A valuation numbers in the billions indicate that AI is moving into a scale-up era where defensibility comes from integrated stacks, not point solutions. Hark and Decart exemplify this trend: both are vertically oriented, both invest heavily in infrastructure for real-time AI inference, and both target broad horizontal platforms rather than single-application products. For founders, this raises the bar—ambitious technical roadmaps, deep hardware partnerships, and clear ecosystem strategies are increasingly necessary to attract top-tier capital. For investors, these rounds show a willingness to underwrite large, long-term bets on platform companies that could anchor entire segments of the AI economy. Taken together, these mega-Series A rounds suggest the market is converging on a thesis that the next wave of value will come from AI that is persistent, multimodal, and embedded directly into the devices, services, and physical systems people already use.