OpenAI Acquires Hiro Finance, Pushing Into Personal Finance Inside ChatGPT

TECH BRIEF

Hello and welcome to Front Research. Here's what's happening in tech today.

• Amazon agrees to buy Globalstar for $11.57B, rolling Apple's iPhone satellite tech into its Leo network and drawing an open battle line with Starlink

• Microsoft hikes Surface PC prices by up to $500 as the global DRAM shortage forces the first visible pass-through to consumer hardware

• Google DeepMind ships Gemini Robotics-ER 1.6, a physical-AI reasoning model that reads factory gauges and orchestrates Boston Dynamics robots

• OpenAI buys personal finance startup Hiro, its latest vertical acquihire as it builds financial planning directly into ChatGPT

Let's jump in.

Amazon Buys Globalstar for $11.57B, Bringing Apple Into Its Starlink Fight

• Amazon announced a $11.57B cash-or-stock deal to acquire Globalstar, paying $90 per share with a 0.3210 Amazon share alternative, expected to close in 2027

• The transaction hands Amazon Leo roughly 24 operational LEO satellites, a planned expansion to 54, ground infrastructure, and scarce licensed mobile-satellite spectrum

• A parallel Apple agreement keeps Emergency SOS, Messages via satellite, Find My, and Roadside Assistance alive on iPhone and Apple Watch, now running on an Amazon-owned network

• FCC Chair Brendan Carr said the agency will review the deal and is "very open-minded," with approval likely to install Amazon as the first credible direct-to-cell challenger to Starlink

• Why it matters: Starlink finally has a fully capitalized, spectrum-rich rival, and Apple's installed base functions as a built-in consumer distribution channel for Amazon from day one

Amazon announced on April 14 a definitive agreement to acquire satellite operator Globalstar for roughly $11.57 billion, paying $90 per share in cash or 0.3210 Amazon shares, with Globalstar holders choosing the mix. The deal is expected to close in 2027 pending regulatory approval and gives Amazon's Project Kuiper low-earth-orbit business, now rebranded Amazon Leo, direct-to-device spectrum, ground infrastructure, and a fleet of roughly 24 operational satellites that Globalstar is already expanding to 54. (CNBC, Bloomberg, TechCrunch)

The surprise inside the deal is Apple. Globalstar powers the Emergency SOS, Messages via satellite, Find My, and Roadside Assistance features on iPhone and Apple Watch, and Apple had taken a 20% stake in Globalstar in 2024 as part of a $1.5 billion investment. Amazon disclosed an accompanying agreement with Apple that preserves and extends those services for current and future iPhones and Watches, meaning Apple's flagship satellite feature set is now effectively running on Amazon infrastructure. (NBC News, GeekWire)

FCC Chair Brendan Carr told CNBC the agency will review the transaction and said he is "very open-minded" about an approval that would install Amazon as a credible direct-to-cell competitor to SpaceX. Musk's Starlink today has roughly 10 million customers and a dominant lead in consumer broadband from orbit, but Amazon is now buying, rather than building, the hardest piece of the stack: scarce, licensed mobile-satellite spectrum. (The Register)

Why it matters: This is the first time Starlink has a fully capitalized, spectrum-rich challenger, and the Apple hook locks a consumer distribution channel in from day one. For Amazon, it transforms Leo from a credible project into a near-term commercial network with an anchor customer. For Apple, it keeps a core safety feature running but hands the underlying network to a direct competitor in cloud and smart home. And for Starlink, which has been carrying the sector's premium valuation, the competitive moat around spectrum just narrowed meaningfully.

Microsoft Raises Surface Prices by up to $500 as the RAM Crunch Hits Consumer PCs

• Microsoft quietly raised Surface PC prices across the entire lineup on April 13 and 14, with hikes of up to $500 per device

• The 13-inch Surface Pro jumped from $999 at 2024 launch to $1,499.99; the 12-inch Surface Pro moved from $799 to $1,049.99; the 15-inch Surface Laptop 7 now starts at $1,600

• Microsoft confirmed the pass-through is driven by the global DRAM and NAND shortage; TrendForce estimates DRAM contract prices rose 90 to 95% in Q1 2026 as hyperscalers front-ran the memory supply for AI data centers

• At the new prices, a base Surface Laptop 7 now costs more than a comparably specced MacBook Air, a reversal already showing up in retail coverage

• Why it matters: This is the first clean consumer signal that AI data-center demand is crowding out non-AI silicon, and it reframes gross-margin guidance across the PC and smartphone supply chain for the quarter

Microsoft quietly raised prices across its entire Surface PC lineup on April 13 and 14, with increases of up to $500 per device. The 12-inch Surface Pro moved from $799 to $1,049.99, the 13-inch Surface Pro jumped from $999 at 2024 launch to $1,499.99, and the 15-inch Surface Laptop 7 now starts at $1,600, up from $1,300 at launch. Microsoft confirmed the hikes are driven by the global DRAM and NAND flash shortage. (Engadget, Windows Central, 9to5Google)

The backdrop is a memory market the industry is now openly calling "RAMageddon." TrendForce estimates DRAM contract prices rose 90 to 95% in Q1 2026 alone as hyperscalers aggressively pre-bought memory for AI data center capacity, leaving consumer and enterprise PC builds fighting for constrained supply. Forecasters expect the squeeze to persist into 2027. (Tom's Guide, Dataconomy)

Microsoft is the first major Windows OEM to pass the full cost through visibly, and at current prices a base Surface Laptop 7 now costs more than a comparably specced MacBook Air, a reversal that is already showing up in retail coverage. (9to5Mac)

Why it matters: This is the first clean consumer-level signal that AI data center demand is crowding out non-AI silicon markets, and it should change how investors read gross-margin guidance across the PC and smartphone supply chain this quarter. Memory vendors Micron, SK Hynix, and Samsung look structurally better positioned, while PC OEMs Dell, HP, and Lenovo will have to choose between margin compression and price-led demand destruction. Apple, which pre-buys memory at longer contract horizons and builds unified DRAM into its M-series silicon, suddenly looks insulated. Watch for similar pass-through announcements from Dell and HP in the next two weeks.

Google DeepMind Ships Gemini Robotics-ER 1.6, a Real Step for Physical AI

• Google DeepMind released Gemini Robotics-ER 1.6 on April 14, available to developers immediately via the Gemini API and Google AI Studio

• The model acts as a high-level reasoning layer for robots, orchestrating vision-language-action models, Google Search calls, and third-party tools across multi-step physical tasks

• On instrument reading (a capability co-developed with Boston Dynamics), the model hits 93% success vs. 23% for version 1.5, a near-fourfold jump that maps directly to industrial inspection and maintenance

• The launch lands one week after SoftBank, Sony, Honda, and NEC formed a trillion-parameter "physical AI" consortium in Japan; Google is shipping a product into the space while the Japanese group is still at company-formation stage

• Why it matters: A model that can reliably read a gauge is the capability industrial customers have been waiting for before signing procurement contracts, and it puts Alphabet back in the frontier-AI narrative

Google DeepMind released Gemini Robotics-ER 1.6 on April 14, making it available to developers immediately through the Gemini API and Google AI Studio. The model functions as a high-level reasoning layer for robots, coordinating vision-language-action models, calling Google Search for information retrieval, and orchestrating third-party tool calls across multi-step physical tasks. Boston Dynamics was a design partner, and DeepMind demonstrated the upgrade driving Spot robots to read analog gauges and sight glasses in industrial environments. (Google DeepMind, Google Blog)

The benchmark move is concrete, not theoretical. On instrument reading, Gemini Robotics-ER 1.6 reaches 93% success versus 23% for version 1.5, a near-fourfold jump in a task class that directly maps to commercial inspection, maintenance, and factory monitoring work. DeepMind also reports substantial gains on spatial reasoning, pointing, counting, success detection, and adherence to physical safety constraints, which are the failure modes that have historically kept embodied AI stuck in demo videos rather than deployed on factory floors. (Seeking Alpha, Blockchain News)

This lands a week after SoftBank, Sony, Honda, and NEC formed a trillion-parameter "physical AI" consortium in Japan targeting the same layer of the stack. Google is now shipping a developer-accessible product into the space while the Japanese group is still at company-formation stage.

Why it matters: Physical AI is the clearest non-chatbot use case for frontier models, and it is also the one where Nvidia, Boston Dynamics, and Google have the clearest distribution advantages. A model that can reliably read a gauge and respond to what it sees is the capability industrial customers have been waiting for before signing procurement contracts. For Alphabet, this is a credible counter to the narrative that Meta's Muse Spark and OpenAI have taken the frontier lead. For Nvidia, robotics-ready Gemini deployments running on Blackwell silicon expand the addressable market on the edge. For the robotics hardware stack (Boston Dynamics, ABB, Fanuc, Teradyne), the bottleneck is moving from mechanical capability to software intelligence.

OpenAI Acquires Hiro Finance, Pushing Into Personal Finance Inside ChatGPT

• OpenAI confirmed on April 14 that it acquired Hiro, an AI-native personal finance startup, in what both sides described as a talent-and-tech acquihire; terms were undisclosed

• Hiro stops new signups on April 20 and deletes all user data on May 13, with an export window before cutoff

• Founder Ethan Bloch previously built neobank Digit, sold to Oportun for more than $200 million in 2021, so OpenAI is acquiring a senior fintech operator along with a financial-reasoning stack that includes a verifiability layer for calculations

• The deal fits an emerging pattern of OpenAI vertical acquihires (Roi in finance, Torch Health in healthcare, now Hiro) paired with new enterprise distribution through the Amazon Bedrock partnership

• Why it matters: The thesis that vertical AI startups would disintermediate horizontal model providers is reversing; OpenAI is now the one buying the vertical, and incumbent wealth managers and robo-advisors are directly in the blast radius

OpenAI confirmed on April 14 that it has acquired Hiro, an AI-native personal finance startup, in what the company and founder Ethan Bloch described as a talent and technology acquihire. Terms were not disclosed. Hiro will stop new signups on April 20 and delete all user data on May 13, with users able to export before the cutoff. (TechCrunch, PYMNTS, Fintech Futures)

Hiro built an AI assistant specifically trained on financial math, with a verifiability layer that let users audit the model's calculations on scenarios like retirement modeling, debt paydown, and taxes. Bloch previously founded neobank Digit, which Oportun acquired in 2021 for more than $200 million, so the deal brings a senior fintech operator into OpenAI alongside a tuned financial reasoning stack. (Unite.AI, Dataconomy)

The deal fits a pattern. OpenAI has spent recent quarters acquiring vertical AI startups in finance (Roi), healthcare (Torch Health), and now personal finance, and pairing those teams with enterprise distribution through the newly announced Amazon Bedrock partnership. (FinancialContent)

Why it matters: The thesis that vertical AI startups would disintermediate horizontal model providers is being quietly reversed: OpenAI is now the one buying the vertical. For incumbent wealth managers and robo-advisors (Wealthfront, Betterment, Charles Schwab, and the advisor tech stack at Envestnet and LPL), an AI assistant that can handle personal financial modeling natively inside ChatGPT is a direct consumer-facing threat. For OpenAI, personal finance is a high-intent, high-retention vertical with clear monetization, and the Hiro verifiability layer is exactly the kind of trust-building scaffolding the category requires. Expect more acquihires in tax, estate, and small-business accounting.