Research note · July 2026
Over the past three years, many of you joined us in securing early positions in Anthropic. What started as an ambitious bet on frontier AI safety has transformed into one of the most remarkable growth trajectories in modern tech, validating the power of acquiring foundational infrastructure early.
As frontier LLMs mature, the next wave of value creation is shifting toward core enabling infrastructure: lower compute costs, real-time inference, high-quality expert training data, and physical robotics. We currently have access to allocations in secondary SPVs for five category-defining companies addressing these critical bottlenecks.
1. Fireworks AI — Next-Gen Model Serving & Inference Cloud
Website: fireworks.ai
Current Round: $1.5B Series D at a $17.5B valuation
Key Investors: Led by Atreides Management, Index Ventures, and TCV; with participation from NVIDIA, Sequoia Capital, Lightspeed, and Bessemer.
Investment Thesis: Founded by PyTorch co-creator Lin Qiao, Fireworks provides high-speed inference backends for enterprise platforms including Perplexity, Cursor, Notion, and Uber. It delivers low-latency model inference and fine-tuning aimed at significantly reducing compute costs compared to traditional hyperscalers.
Funding Round: fireworks.ai/blog/series-d-announcement
2. Mercor — The Human Intelligence Engine for AI
Website: mercor.com
Current Round: Series C / Extension (seeking $500M+ at a $20B valuation)
Key Investors: Felicis Ventures, Benchmark, General Catalyst, and Menlo Ventures.
Investment Thesis: As AI models exhaust synthetic data, Mercor provides specialized human expert pipelines required to train, evaluate, and benchmark frontier models. Supporting data workflows for leading AI labs including OpenAI, Anthropic, and Meta, Mercor is positioning itself as an essential player in high-end AI data infrastructure.
Funding Round: aiweekly.co
3. Positron AI — Custom Silicon Challenging the GPU Bottleneck
Website: positron.ai
Current Round: $230M Series B at a $1B+ valuation
Key Investors: Co-led by ARENA, Jump Trading, and Unless; with strategic backing from Arm Holdings, QIA, and Valor Equity Partners.
Investment Thesis: Built specifically to address energy and cost constraints in LLM inference, Positron designs specialized silicon hardware intended to bypass traditional GPUs. Backed by Arm and major quantitative trading firms, Positron aims to lower the global AI compute cost curve over time.
Funding Round: af.net
4. Atoms — Industrial AI & Physical Automation
Website: atoms.co
Current Round: $1.7B equity & debt round at a $15B valuation
Key Investors: Led by Andreessen Horowitz (a16z); with participation from Bain Capital, Fifth Wall, K5 Global, and Uber.
Investment Thesis: Founded by Travis Kalanick (co-founder of Uber), Atoms applies advanced compute, AI, and industrial robotics to heavy physical industries including mining, construction, and transportation. With Ben Horowitz joining the board, Atoms represents a potential market leader in the transition from digital AI (“bits”) to real-world deployment (“atoms”).
Funding Round: kucoin.com
5. Together AI — Open-Source Cloud Infrastructure & Compute
Website: together.ai
Current Round: $800M Series C at an $8.3B valuation
Key Investors: Led by Aramco Ventures; with participation from NVIDIA, Vista Equity Partners, General Catalyst, Emergence Capital, and S Ventures.
Investment Thesis: Together AI provides high-performance cloud infrastructure allowing enterprises to run, train, and scale open-source AI models efficiently. As companies look to optimize margins, Together AI offers critical infrastructure aimed at dramatically reducing inference costs compared to proprietary models.
Funding Round: techfundingnews.com
If you would like to receive more information on any of these opportunities, including term sheets and data rooms where available, please reach out directly.