Milk Road AI
Intelligence extracted from Milk Road AI newsletters.
13
Issues Tracked
30
Insights Extracted
6
Topics Covered
Topics
Key Insights from Milk Road AI
**Cerebras Systems** lists on Nasdaq as CBRS with its IPO 20x oversubscribed, powered by its WSE-3 wafer-scale chip that is 56x larger than **Nvidia's H100** and delivers ~7,000x faster memory bandwidth.
The AI inference market is projected to grow from $106B in 2025 to $255B by 2030, and **Cerebras** benefits from rising HBM memory costs making Nvidia-based infrastructure more expensive while its on-chip SRAM architecture sidesteps that supply chain entirely.
Key risks include 86% of 2025 revenue from two UAE-tied entities, a conditional $20B **OpenAI** contract that can be terminated, and a governance conflict involving **Sam Altman** as personal investor, customer CEO, and creditor simultaneously.
Author rotated from **Robinhood** to **Oracle** after HOOD missed Q1 2026 revenue consensus by 13% due to crypto transaction revenue cratering 47% YoY, while the stock still trades near peak-narrative valuations.
**Oracle** reported 84% IaaS growth for the fifth consecutive accelerating quarter and holds $553B in remaining performance obligations—~$300B from **OpenAI** alone—with FY2027 guidance raised to $90B.
**Anthropic** sealed a deal with **SpaceX**, which is also exploring putting AI data centers in orbit according to an **ARK Invest** interview on the Milk Road AI Show.
**Amazon Supply Chain Services (ASCS)** opens Amazon's 13B-package-per-year logistics network — ocean freight, warehousing, last-mile delivery, customs — to any business with no long-term contracts, mirroring the AWS playbook.
**Amazon's Q1 2026 earnings** hit $181.5B revenue (+17%), $30.3B net income (+77%), and AWS contracted backlog surged $120B in 90 days to $364B, funded by $200B in 2026 CapEx.
**FedEx, UPS, Shopify Fulfillment, Flexport, and ShipBob** face existential pressure as Amazon targets the $1.4T global 3PL market, which is 3x the size of the cloud market when AWS launched.
**Amazon** reached $15B AI revenue run rate, the fastest commercial technology adoption in company history, driven by **Trainium** chips, **Bedrock** platform, and **SageMaker**.
Latest issue: May 13, 2026
🥛 The most important AI IPO? 👀
Cerebras Systems is listing on Nasdaq under ticker CBRS on May 14, 2026, with its IPO 20x oversubscribed, driven by its Wafer-Scale Engine chip that is 56x larger than Nvidia's H100 and 7,000x faster in memory bandwidth. The newsletter argues the AI industry is shifting from training to inference, a market projected to grow from $106B to $255B by 2030, positioning Cerebras favorably. However, significant risks include 86% revenue concentration from UAE-tied entities, a conditional $20B OpenAI contract, and a complex governance structure involving Sam Altman.
🥛 The trade I’m making right now 👀
The author explains their decision to rotate out of Robinhood stock into Oracle, citing Robinhood's over-reliance on volatile crypto transaction revenue and a 13% Q1 2026 earnings miss. Oracle is positioned as a contrarian AI infrastructure play with $553B in remaining performance obligations, 84% IaaS growth, and a projected cash flow inflection around FY2029. The piece argues Oracle's narrative lag—still perceived as a legacy database company—creates asymmetric upside.
🥛 The scariest Amazon launch yet 😳
Amazon launched Amazon Supply Chain Services (ASCS), opening its 13B-item-per-year logistics network to any business, mirroring its AWS strategy of monetizing internal infrastructure. Amazon's Q1 2026 earnings showed $181.5B in revenue (+17%), $30.3B net income (+77%), and AWS growing 28% with a contracted backlog jumping $120B in 90 days to $364B. The move threatens FedEx, UPS, Shopify Fulfillment, and logistics startups by offering end-to-end supply chain services with no long-term contracts.
🥛 Everyone’s buying AI wrong 🤯
🥛 OpenAI just broke the AI deal 🧨
🥛 The AI market is wrong 🤯
🥛 The real AI winner 🧠
Amazon's AI revenue reached $15B annual run rate, the fastest commercial adoption in Amazon's history, while the company plans $200B in capex for 2026. Amazon's custom chips (Graviton, Trainium, Nitro) generate over $20B annually and may be sold to third parties as a standalone business worth $50B.
🥛 The AI advantage you can’t copy 🧠
Meta is executing a three-part AI strategy involving acquiring Scale AI for $14.3B, buying autonomous AI company Manus for $2.5B, and integrating AI into Ray-Ban glasses that sold 7M units in 2025. However, the strategy faces challenges as founding AI scientist Yann LeCun left to start AMI Labs with $1.03B funding, and Meta's new flagship model 'Avocado' was delayed due to performance issues.
🥛 The AI Bubble just got real 🚨
OpenAI closed a $122B funding round at $852B valuation from Amazon, Nvidia, and others, then acquired tech livestream TBPN for over $150M while simultaneously killing their Sora video product due to unsustainable GPU costs. The company faces compute constraints despite massive funding, with daily GPU rationing meetings and upcoming IPO pressure alongside SpaceX and Anthropic potentially flooding public markets.
🥛 The hidden robotaxi winner 🏆
Uber is positioned as the hidden winner in the robotaxi war by becoming the essential platform that connects autonomous vehicle companies with riders, similar to how Visa operates in payments. While Tesla and Waymo battle for technical supremacy with vertical integration strategies, Uber has built partnerships with 20+ AV companies across four continents and launched Uber Autonomous Solutions to license its infrastructure.
🥛This AI selloff makes no sense❓
Google's TurboQuant compression algorithm triggered an AI memory stock selloff, but the technology only compresses one component (KV cache) by 6x while leaving three other memory types untouched. The efficiency gains will likely unlock new AI applications with longer context windows, following historical patterns where efficiency improvements increase rather than decrease total demand.
🥛 The most dangerous bet in AI ⚠️
Elon Musk is attempting to replicate Andrew Carnegie's vertical integration strategy in the semiconductor industry with Terafab, a joint Tesla-SpaceX-xAI chip mega-facility targeting 1 terawatt of AI compute annually. The facility aims to produce logic chips, memory chips, and space-hardened inference chips under one roof, with 80% of output destined for SpaceX's planned constellation of up to 1 million AI satellites in low Earth orbit. The single critical bottleneck is ASML, the sole provider of EUV lithography machines, whose order book is full through 2028 and with whom Tesla has no existing relationship.
🥛 The AI trade nobody sees 👀
Milk Road AI argues that optical transceivers — not GPUs — are the overlooked infrastructure layer powering the AI buildout, with NVIDIA's $4B investments in Coherent Corp and Lumentum signaling an impending supply crunch in InP laser fabrication. The newsletter profiles four public companies (COHR, LITE, AAOI, CRDO, MRVL) positioned to benefit from the 36x increase in fiber demand that AI data centers require versus traditional CPU-based systems. Jensen Huang's doubling of AI infrastructure demand projections to $1T through 2027 implies $50B in cumulative optical networking spend that most analyst models are underestimating by 2x.