Two Funding Rounds, Over $1B, One Clear Signal
In the span of days, two AI startup funding rounds have pushed more than USD 1 billion (approx. RM4.6 billion) into foundational AI infrastructure. Hark raised over USD 700 million (approx. RM3.2 billion) in an oversubscribed Series A at a USD 6 billion (approx. RM27.6 billion) series A funding valuation, while Decart secured USD 300 million (approx. RM1.4 billion) to scale its real-time AI inference stack. Both companies are building deep-technology platforms rather than narrow applications, and their cap tables are stacked with strategic investors such as Nvidia and other major venture firms. The combined scale of these AI startup funding rounds underscores that investors are still aggressively backing capital-intensive research labs, especially those that own key parts of the compute, model, and hardware pipeline. Rather than signaling a cooling market, these deals suggest a rotation toward infrastructure-heavy bets that can power the next generation of AI-native products.
Hark: Building a Universal Human–Machine Interface
Hark’s oversubscribed Series A is notable not just for its size but for its ambition. The company is explicitly avoiding a single-layer play in the AI stack, instead combining models, software, and AI-native hardware into a unified platform. Its vision is a “universal interface between humans and machines” that goes beyond chatbots to deeply personalized, context-retaining assistants that “actually know you” across services and devices. With roughly 70 employees and a new NVIDIA B200 data center for training, Hark is positioning itself as both an AI systems lab and an AI hardware development company. The planned release of its first agentic, multimodal models later this summer will be an early test of whether tightly integrating custom hardware with personalized AI experiences delivers a defensible advantage—or simply raises the bar for capital requirements in this category.
Decart: Real-Time AI Inference for Video and the Physical World
Decart is betting that the future of AI will be live, continuous, and deeply embedded in visual environments. The vertically integrated lab focuses on real-time video, world models, and a next-generation inference stack designed for speed. Its core product, DOS 2.0, reportedly delivers over 1,600 tokens per second for agentic, real-time AI inference—around eight times the industry average—plus full-HD video at up to 100 frames per second across major cloud hardware providers. On top of this stack, Decart offers two model lines: Lucy, targeting real-time immersive experiences in gaming, e-commerce, and advertising with sub-30-millisecond responses; and Oasis, aimed at robotics and autonomous vehicles via physically accurate, real-time simulation. Together, they illustrate how AI infrastructure is evolving from static model serving toward high-throughput, low-latency systems built for interactive and physical-world applications.
What These Mega Rounds Reveal About Investor Priorities
Taken together, Hark and Decart’s raises show investors gravitating toward startups that control critical AI infrastructure and hardware layers rather than pure software wrappers. Hark’s focus on custom devices as a universal interface and Decart’s DOS 2.0 inference stack both reflect a belief that performance, latency, and tight hardware–software integration will be key moats. The investor mix—featuring leading chipmakers, cloud-aligned venture arms, and top-tier funds—suggests that strategic access to compute and deployment channels is now as important as model quality. These AI startup funding rounds also highlight a renewed appetite for AI hardware development and real-time AI inference capabilities that can support agentic systems, world models, and immersive applications. For founders and incumbents alike, the message is clear: owning or deeply optimizing the underlying infrastructure is increasingly where the biggest capital and confidence are flowing.