Palantir Revenue Surges 96% to $2.31B, U.S. Commercial Tops Government

Key Takeaways

• <ul><li>Palantir raised its 2026 revenue forecast to $8.53 billion–$8.534 billion after Q3 revenue jumped 96% to $2.31 billion</li><li>U.S. commercial revenue surpassed U.S. government revenue for the first time, reaching $984 million vs. $953 million</li><li>MediaTek targets $16 billion in AI accelerator revenue by 2027, aiming for 20% of the ASIC market</li><li>Fairfax County schools propose banning AI through sixth grade and allowing limited use in middle school with parental approval</li><li>Western startups like Hirundo and Lazarus AI are modifying DeepSeek and other Chinese models to match ChatGPT and Claude safety standards</li><li>Researchers found AI student models can absorb hidden preferences from judge scoring during training</li><li>Domain-specific models BioBERT and BioMistral showed promise in predicting clinical trial outcomes across 12,477 trials</li><li>Midstream Health raised $20M to use AI helping hospitals recover uncollected revenue from contracts</li><li>Highland Park extended its Placer.ai contract through 2029 for $82,145 annually to track foot traffic and business patterns</li><li>Quantum X Labs demonstrated information-theoretic encryption on a 156-qubit IBM Quantum Heron processor</li></ul>

Palantir raises 2026 revenue forecast as AI demand drives near-doubling of quarterly sales

Palantir raised its full-year 2026 revenue forecast to between $8.53 billion and $8.534 billion after third-quarter revenue jumped 96% to $2.31 billion. U.S. commercial revenue reached $984 million, surpassing U.S. government revenue of $953 million for the first time. The company also raised its U.S. commercial revenue growth forecast to at least 143%. CEO Alex Karp highlighted growing adoption of sovereign AI through platforms like Foundry and the Artificial Intelligence Platform.

MediaTek Q3 revenue climbs 18.3% as AI chip sales target $16 billion in 2027

MediaTek reported Q3 2026 revenue of NT$168.06 billion, up 18.3% year over year, driven by strong demand for AI-related chips. The company targets approximately $16 billion in AI accelerator revenue in 2027, about 20% of the AI accelerator ASIC market. Smart Edge Platforms accounted for 58% of revenue and grew 63% year over year. MediaTek began initial shipments of its first AI accelerator in September 2026 and expects $2.5 billion in revenue in Q4 2026.

Fairfax schools draft new policies on AI and classroom technology use

Fairfax County Public Schools presented early drafts of two policies on emerging technologies and educational technology. The proposed rules suggest banning AI use entirely through sixth grade, allowing limited curriculum-grounded tools in grades 7 and 8 with parental approval, and permitting broader AI use in high school with teacher guidance. A formal vote is expected on Feb. 11, 2027. The policies also address student privacy, data security, device usage by grade level, and AI literacy instruction.

Federal IT success should focus on agility not just spending, report says

A new commentary argues that federal IT programs should be measured by agility and mission outcomes rather than activity metrics like tickets closed. Agencies need flexibility to adapt technology as mission needs change, especially with growing investment in AI. The author urges agencies to retain ownership of modernization roadmaps and tie technology investments to measurable improvements in service delivery. Examples include reducing application review times and improving status visibility for grant programs.

Domain-specific AI language models show promise in predicting clinical trial outcomes

Researchers evaluated two domain-specific language models, BioBERT with 108M parameters and BioMistral with 7B parameters, for predicting clinical trial outcomes. The study used the Hierarchical Interaction Network approach to model interactions between drugs, diseases and protocols. Testing covered 12,477 trials on the TOP benchmark. The research aims to reduce costly drug development failures and speed up patient access to effective therapies.

Student AI models absorb hidden traits from judge scoring

Researchers found that AI student models learn hidden preferences from judges during training. A student model learned to prefer certain animals or patterns based only on a judge's scoring style. Of 48 test cases, 17 showed clear trait transfer at a 5% false-discovery rate. The study raises concerns about AI safety and how unwanted traits can pass to student models.

Western startups reshape Chinese AI models for safety

A growing number of startups are modifying Chinese open-weight AI models for Western businesses. Companies like Hirundo and Lazarus AI edit models from DeepSeek, Moonshot, Z.AI, and Alibaba to remove censorship and security flaws. CrowdStrike found Alibaba's Qwen models avoid politically sensitive topics. The realigned models aim to match the safety standards of ChatGPT and Claude. However, experts say making models universally safe remains an impossible standard.

Quantum X Labs tests encryption on IBM quantum processor

Quantum X Labs demonstrated its Qatacomb encryption scheme on an IBM Quantum Heron r2 processor with 156 qubits. The intended recipient successfully recovered an encrypted image by reading the machine's noise. An adversary without the key gained no usable information. The scheme is designed to be information-theoretic, meaning its security does not depend on assumptions about an adversary's computing power.

Midstream Health raises $20M to help hospitals recover revenue

Midstream Health raised $20 million in series A funding led by 7wire Ventures and Town Hall Ventures. The company uses AI to analyze hospital contracts and transactions to find money that is earned but not collected. CEO Sumit Kadakia said health systems are losing margins they have already earned. The funding will expand the technology and hire more staff for growth.

Highland Park extends contract with Placer.ai through 2029

The City of Highland Park renewed its contract with data analytics company Placer.ai for three more years. The new agreement runs from December 1 through November 30, 2029, and costs $82,145 annually. The platform helps the city track foot traffic, business patterns, and event attendance. Local businesses can also request unique data reports to help with decisions about hours and operations.

Sources

• Palantir raises 2026 revenue outlook after quarterly revenue nearly doubles on AI
• MediaTek Q3 2026 Revenue Rises 18.3% to NT$168 Billion, AI Chip Sales Target $16 Billion in 2027
• New FCPS policies governing educational technology, use of AI start to take shape
• Federal IT needs a new measure of success: Agility
• Advancing Clinical Trial Outcome Prediction with Domain-Specific Large Language Models
• Subliminal Reinforcement Learning: traits travel through a judge's scores
• Startups recalibrate Chinese AI for the West
• Quantum X Labs Announces Cyber Quantum Encryption Breakthrough in Demonstration with IBM Quantum Processor, Heron r2
• Midstream Health nabs $20M series A to help hospitals recover lost margin
• Highland Park reups contract with location-data platform Placer.ai