Tech news in 3 minutes

Spectro Cloud Raises $100 Million Series D to Accelerate Production AI Adoption

4 h ago

General Compute, a fast inference neocloud building an ASIC-based alternative to GPU clouds, has secured a committed debt facility of up to $400 million from Upper90 Capital Management to scale one of the world's largest inference neoclouds, delivering AI results 16x faster than current GPUs. The financing starts at $100 million and scales with customer demand, marking one of the largest debt deals for a non-GPU cloud provider. General Compute holds no GPU allocation, freeing it to deploy specialized inference silicon from SambaNova (SN40 and SN50) while GPU clouds remain tied to a single chip supplier. The company has over $300 million in secured, price-protected supply and expects to be the first neocloud to deploy ASIC silicon at scale. Upper90, also an equity investor, aligns debt and equity behind the differentiated approach. The announcement addresses the growing GPU dilemma: as Goldman Sachs predicts token consumption will rise 24x in 3.5 years, GPU clouds prove inefficient for inference, with new GPU racks demanding 120kW of power and specialized liquid cooling, trapping capacity behind multi-year data center builds. General Compute’s solution uses air-cooled ASIC chips requiring only 20kW per rack, no water cooling, and can be installed in weeks instead of years. Key performance claims include 16x faster inference, 7x faster time-to-first-token, 8.5x higher output throughput (1,000 tokens/second), and 6x better power efficiency than GPUs. The platform supports frontier models from OpenAI, DeepSeek, MiniMax, and others, with a switchover time of under 30 seconds. “We are the only neocloud that can serve premium tokens: frontier-level intelligence on the largest models, served fast,” said CEO Finn Puklowski, emphasizing the company’s lack of handcuffs to a single chip supplier.