Google is pushing deeper into custom AI hardware with two new chips: the TPU 8tx, built in partnership with AMD, and the TPU 8i Max, which packs two inference-focused chips into a single package. The TPU 8tx uses AMD's fifth-generation Epyc processor and hardware encryption to protect AI models and data during processing. Together, the chips reflect Google's effort to reduce its reliance on outside suppliers while strengthening security for cloud-based AI workloads.
Google Cloud also made its Pathways Runtime generally available on Google Kubernetes Engine as part of the AI Hypercomputer. The distributed runtime is designed for large AI training and inference jobs. In the past 90 days, companies processed over 130 million Cloud TPU chip-hours using Pathways, which relies on a single-controller architecture to simplify management of complex workloads.
In the earnings space, The Trade Desk delivered a strong Q3 beat, reporting revenue of $736 million against analyst estimates of $651.4 million. Adjusted EPS came in at $0.33 versus consensus of $0.25, and Q4 guidance of $755 million topped expectations. CEO Jeff Green highlighted how AI is adding complexity and data for advertisers, and the company's new Audience Unlimited product uses AI and third-party data to improve targeting. Early tests showed customer acquisition costs falling over 30% and ad costs dropping 14%.
Meanwhile, SoundHound AI reported weaker-than-expected Q3 2026 revenue of $62.43 million, missing estimates of $65.79 million. While revenue grew 48.5% year over year, the company posted a GAAP net loss of $46.4 million and expects adjusted EBITDA profitability in the second half of 2027. The stock has fallen 61.6% since the start of the year.
On the infrastructure front, Microsoft is lobbying the Trump administration against a proposed ban on Chinese-made optical transporters used in AI data centers. The company argues that blocking these components could slow new data center construction and hurt U.S. competitiveness in AI. Microsoft currently has 35 data centers under construction in the United States and another 134 planned.
Financial institutions are moving beyond AI pilots into production deployments, though only about one-third of organizations have the governance and security needed for scaled operations. Companies like New York Life Insurance and Fifth Third Bank are using AI to boost employee productivity and customer service, with cloud modernization serving as the foundation for these efforts.
NVT is expanding its power and cooling product line to meet growing demand from AI data centers. Its modular liquid cooling platform is taking customer orders, and the acquisition of Maverick Power adds power distribution products and an 800VDC power rack designed for high-density AI computing. New products contributed roughly 30 percentage points to NVT's sales growth in Q3, and the company launched 24 new products during the quarter.
TSMC's October sales surged 59.7% year over year to NT$586.9 billion, marking the second fastest rise in 2026. The monthly figure was 17.8% above the third quarter average and represented 34.6% of the company's fourth quarter guidance midpoint. CFO Wendell Huang expects strong performance into Q4 2026, though the report does not specify which products drove the growth.
Broadcom stock fell over 20% in one month amid concerns about AI hardware market exposure. Some analysts view the decline as overdone, pointing to Broadcom's diversified portfolio and strong fundamentals as supporting factors. Investors exploring AI opportunities during the market dip are showing interest in technologies that improve GPU operating efficiency and reduce training costs.
Church Law & Tax and Gloo released an updated 2.0 AI Policy Template that expands guidance for organizations using AI. The new version covers embedded AI features and agentic AI, adds a more robust risk-review process, and introduces Protected Data Terms for tools handling sensitive information. It also addresses AI-assisted hiring decisions, deepfakes, fraud, and verification of unusual requests involving payments or data transfers.
Key Takeaways
- <ul><li>Google launched TPU 8tx with AMD and TPU 8i Max, its eighth-generation AI chips focused on security and inference performance.</li><li>Google Cloud's Pathways Runtime is now generally available, with over 130 million Cloud TPU chip-hours processed in 90 days.</li><li>The Trade Desk beat Q3 expectations with $736 million in revenue versus $651.4 million estimated, and Q4 guidance of $755 million.</li><li>SoundHound AI missed Q3 revenue estimates of $65.79 million with $62.43 million reported, and expects EBITDA profitability in H2 2027.</li><li>Microsoft is lobbying against a proposed ban on Chinese optical transceivers, citing 35 data centers under construction and 134 more planned in the U.S.</li><li>Only about one-third of financial institutions have the governance and security needed for production-scale AI deployments.</li><li>NVT's new power and cooling products contributed approximately 30 percentage points to Q3 sales growth, with 24 new products launched.</li><li>TSMC's October sales rose 59.7% year over year to NT$586.9 billion, the second fastest rise in 2026.</li><li>Broadcom stock fell over 20% in one month due to AI hardware market exposure, though some analysts see it as undervalued.</li><li>The updated 2.0 AI Policy Template expands coverage to agentic AI, embedded AI features, hiring decisions, and deepfake fraud prevention.</li></ul>
AMD Epyc powers Google Cloud TPU 8tx with hardware encryption
AMD's fifth-generation Epyc processor is now built into Google Cloud's TPU 8tx AI accelerator. The system uses hardware-based confidential computing to protect data while it is being processed. AMD's SEV technology encrypts AI models, user prompts, and data moving between the CPU and TPU. This setup helps protect sensitive information on cloud infrastructure.
Google announces two new AI chips TPU 8tx and TPU 8i Max
Google introduced two new specialized chips, TPU 8tx and TPU 8i Max, as part of its eighth-generation hardware lineup. TPU 8tx, developed with AMD, focuses on securely training and running AI models. TPU 8i Max combines two inference-focused TPU 8i chips into one package with double the memory and processing power. The launch aims to reduce Google's reliance on outside chip suppliers.
Google Cloud releases Pathways Runtime for AI training and inference
Google Cloud announced the general availability of Cloud Pathways on Google Kubernetes Engine as part of the Google Cloud AI Hypercomputer. Pathways is a distributed runtime designed for large AI workloads using many accelerators. In the past 90 days, companies processed over 130 million Cloud TPU chip-hours using Pathways. The system uses a single-controller architecture to simplify management of large AI training jobs.
Trade Desk stock rises after Q3 beat with AI-driven growth plans
The Trade Desk beat Q3 earnings expectations and reported strong Q4 guidance. CEO Jeff Green said AI is creating more data, choices, and complexity for advertisers. The company launched Audience Unlimited to use AI and third-party data for better audience targeting. Early tests showed customer acquisition costs dropping over 30% and ad costs falling 14%.
Trade Desk Q3 2026 results show strong guidance and AI progress
The Trade Desk reported Q3 2026 revenue of $736 million, beating analyst estimates of $651.4 million. Adjusted EPS reached $0.33 versus consensus of $0.25. Q4 revenue guidance came in at $755 million, above expectations. The company highlighted AI integration through its Zuma platform, joint business plans, and strong growth in Asia-Pacific markets.
Financial institutions tackle scaling AI responsibly
Financial institutions are moving beyond pilot projects to deploy AI at scale across regulated operations. Only about one-third of organizations have the governance and security needed for production AI. Banks and insurers must meet strict requirements for auditability and customer trust. Companies like New York Life Insurance and Fifth Third Bank are using AI to improve employee productivity and customer service. Cloud modernization provides the foundation for these enterprise AI deployments.
Investors explore AI stock opportunities during market dip
Even during a market correction, investors are buying into companies that improve GPU operating efficiency. These technologies reduce training costs and increase equipment utilization for AI operations. Stock-backed loans are gaining popularity as a way to access funds without selling shares. All New Stock Loan launched a product with interest rates around 5% per year. Investors can borrow up to four times their capital and use the loans for share purchases or refinancing.
TSMC October sales surge 59.7% to record high
TSMC reported October consolidated sales of NT$586.9 billion, a 59.7% year-on-year increase and the second fastest rise in 2026. October represented 34.6% of the company's fourth quarter guidance midpoint. The monthly figure was 17.8% above the third quarter average. TSMC's CFO Wendell Huang expects strong performance into fourth quarter 2026. The report provides an early read on AI chip demand but does not specify which products drove the growth.
Broadcom stock tested by AI growth and valuation concerns
Broadcom stock fell over 20% in one month due to its exposure to the AI hardware market downturn. Investors are reassessing the company's valuation and growth prospects. Some analysts believe the stock is undervalued and expect a long-term recovery. Broadcom's diversified product portfolio and strong fundamentals are seen as supporting factors. The current AI hardware weakness is viewed by some as a temporary setback.
SoundHound AI stock drops on weak revenue report
SoundHound AI shares fell 5.1% after reporting third quarter 2026 revenue of $62.43 million, below analyst estimates of $65.79 million. Revenue grew 48.5% year over year. The company reported a GAAP net loss of $46.4 million and expects adjusted EBITDA profitability in the second half of 2027. The stock is down 61.6% since the start of the year and trading at $4.08 per share.
NVT expands power cooling products for AI data centers
NVT is expanding its portfolio of power and cooling products to address growing demand. Its modular liquid cooling platform has started receiving customer orders and serves different types of data center customers. The company is also developing solutions for higher-voltage power systems. During the third-quarter earnings call, management highlighted that its new products contributed approximately 30 percentage points to sales growth. NVT also launched 24 new products during the quarter, with a focus on high-growth markets, such as data centers and power utilities. The acquisition of Maverick Power further expands NVT's exposure to AI data centers, bringing power distribution products and an 800VDC power rack designed for high-density AI computing.
Microsoft lobbies US against ban on Chinese AI hardware
Microsoft has urged the Trump administration to halt a proposed ban on U.S. imports of certain Chinese-made optical transceivers used in AI data centers. Industry groups warn that blocking these components could severely slow new data center buildouts, undermining U.S. competitiveness in artificial intelligence. Microsoft has 35 data centers under construction across the United States and another 134 planned. The Federal Communications Commission is working on the proposed ban, while lawmakers have separately introduced legislation targeting Chinese optical transceivers in sensitive government systems. Microsoft and other companies argue an import ban could inflict lasting harm on the U.S. lead over China in AI by further slowing the deployment of new data centers.
How to update AI policy to address new risks
Church Law & Tax and Gloo have released an updated 2.0 AI Policy Template. The first template provided churches with a framework for approving AI tools and establishing acceptable uses. Version 2.0 goes further by expanding coverage to embedded AI features and agentic AI. It establishes a more robust AI tool and provider risk-review process, including data retention and security safeguards. The template introduces Protected Data Terms for AI tools that may handle sensitive information. It adds safeguards for AI recording, transcription, and note-taking involving pastoral care and counseling. The policy broadens protections involving AI-assisted hiring, compensation, promotion, and other employment decisions. It also addresses AI-generated impersonation, deepfakes, and fraud, including verification of unusual requests involving payments or data transfers.
Sources
- AMD Epyc to power Google Cloud TPU 8tx with hardware confidential computing
- Google broadens its AI hardware portfolio with two new specialized chips
- Google Cloud Releases Pathways Runtime for Large-Scale AI Training and Inference
- TTD Stock Rises Overnight After Q3 Beat: Trade Desk CEO Says AI Creates ‘More Data, More Choices And More Complexity’
- TTD Q3 2026 deep dive: operational discipline, AI-driven products, and guidance surprise
- The challenge of scaling AI in financial services
- AI Investment Landscape Shifts... Is "Now" the Time to Buy the Dip?
- TSMC's October Sales Jump 59.7% to a 2026 High of NT$586.9B
- Broadcom stock faces an AI growth test as valuation resets
- Why SoundHound AI (SOUN) stock is down today
- Can NVT's AI Data Center Exposure Accelerate its Revenue Growth?
- Microsoft Lobbies US Against Ban on Chinese AI Hardware
- How To Update AI Policy To Address New Risks From Agentic AI, Deepfakes, and Data Security
Comments
Please log in to post a comment.