Maya Scores 96.7% Ahead of Google DeepSeek and OpenAI

SquadStack.ai launched Maya, a model for live sales calls, on Nov. 16, 2026. The company says Maya scored 96.7% on a benchmark built from 9,962 moments in 471 Indian sales calls, ahead of Google’s Gemma 4 31B (92.5%), DeepSeek V4.1 Flash (92.2%) and OpenAI’s GPT-6 Luna (89%). Anthropic’s Claude Haiku 5.5 judged benchmark replies. SquadStack says it trained Maya on more than 40 crore sales interactions collected over six years; it handles over 50 lakh conversations daily in Hinglish and other Indian languages, starts replies in a median 278 milliseconds, and is hosted in India. It runs within the company’s Voice AI Agents, while its Lift system learns from call results. Maya is in closed beta with a small group of Indian sales teams, and SquadStack is exploring live cues such as pauses and tone.

Two articles identify Jabil and Applied Digital as potential beneficiaries of AI infrastructure demand, alongside suppliers such as Nvidia and Micron. Jabil’s fiscal 2026 revenue rose 21% to $36 billion; it forecasts $44.5 billion in fiscal 2027 revenue, including $22.1 billion from AI, up 54%. Applied Digital reported fiscal 2027 first-quarter revenue of $342 million, up 322% year over year. It plans to deliver another 600 megawatts of data-center capacity over the next 12 months and targets 3.5 to 4 gigawatts by the end of 2030. Its contracted lease pipeline totals $36 billion, tied to 1.4 gigawatts. The second article offers no additional company figures.

Researchers studying two commercial text-to-video platforms estimate emissions could climb nearly 75-fold by 2030 if video generation continues growing, even with efficiency gains. If usage instead remains at 2026 levels, those improvements could reduce emissions by about two-fifths. The researchers describe the estimates as a conservative lower bound because they cover only one AI service among many.

AhnLab said Nov. 16 that subsidiary AhnLab Cloudmate won the Korea Startup Forum Chairman’s Award at the 21st Korea Internet Awards, held in Seoul on Nov. 13. The award recognized SecureBridge, which uses patented context analysis to detect and block personal or confidential information in prompts without extra software agents. The tool has standard-level SaaS certification under South Korea’s Cloud Security Assurance Program and was named a 2026 outstanding information security technology by the Ministry of Science and ICT and KISA.

Anadyr Horizon co-founder and CEO Arvid Bell discussed the company’s use of large-scale AI simulations for geopolitical and investment risk analysis. Separately, a Moody’s Ratings report says U.S. life insurers held $85.2 billion in AI debt as of June 30, up 24% in six months and equal to 2.1% of their total debt. Moody’s says holdings could reach $185 billion by the end of next year. The report notes that a financial entity created by Elon Musk and advisers borrowed billions to buy AI chips and lease them to xAI, keeping the debt off xAI’s official books; insurer Athene bought $2.4 billion of it.

Uber says its Legal Redlining Agent has reduced average contract-review time by more than 20%, with 91% accuracy in AI-generated decisions. Integrated into Microsoft Word, it reviews contract changes, checks risks and policies, and recommends whether lawyers should accept, reject or revise clauses. It combines retrieval-augmented generation, past lawyer feedback and fixed rules; lawyers still review and approve recommendations. Uber began development in early 2024 and piloted the tool with its Legal team in 2025.

Wealth-management firms are also expanding AI offerings. Osaic approved Nitrogen’s six-product adviser suite, including the Nucleus AI agent, which can automate tasks such as tax-return analysis. Masttro introduced a Model Context Protocol server and developer portal for eligible clients to connect wealth data with third-party AI applications; read-only access began Nov. 16. Glide launched a U.S. portfolio-analysis tool that estimates asset allocation, returns and risk, with availability for investors and advisers starting Nov. 30.

Analysts differ on whether heavy AI infrastructure spending poses a broader risk, though some view the recent AI-stock sell-off as a repricing rather than a likely crash. Generali’s Vladimir Oleinikov said cloud-company revenue forecasts for 2026–2028 rose by $235 billion while free-cash-flow estimates fell by $163 billion. Morgan Stanley called the sell-off an overreaction and expects hyperscalers to spend $1.6 trillion in 2028 and $1.8 trillion in 2029, with AI returns on invested capital above 30%. TS Lombard’s Dario Perkins warned that growth depends heavily on spending within the sector rather than outside customers. About 20% of expected 2028 data-center capacity is delayed until 2029.

University of Glasgow researcher Dr. Roisin Buckley received funding through a £39 million UKRI initiative to develop AI tools for offshore-wind seabed and subsurface surveys. Her team will combine geological, geophysical and engineering data with physics, geology and engineering expertise to identify information gaps and communicate uncertainty in predictions. The work aims to reduce investigations that can take around 10 years and cost tens of millions of pounds. With Scottish Power Renewables, the team has assembled a curated subsurface-data database; the approach could also support transport, telecommunications, nuclear energy and carbon-capture projects.

Key Takeaways

  • SquadStack.ai says its India-hosted sales model Maya scored 96.7% on a benchmark, ahead of Google, DeepSeek and OpenAI models; it handles over 50 lakh conversations daily.
  • Jabil forecasts fiscal 2027 AI revenue of $22.1 billion, up 54%; Applied Digital reports $342 million in quarterly revenue and targets 3.5–4 gigawatts of data-center capacity by 2030.
  • Researchers estimate text-to-video emissions could rise nearly 75-fold by 2030 if generation grows, while holding usage at 2026 levels could cut emissions by about two-fifths through efficiency gains.
  • AhnLab Cloudmate’s SecureBridge won a Korea Internet Awards honor for protecting sensitive information in generative-AI prompts.
  • Moody’s says U.S. life insurers held $85.2 billion in AI debt as of June 30, up 24% in six months; it projects holdings could reach $185 billion by the end of next year.
  • Uber says its Microsoft Word-based Legal Redlining Agent cuts average contract-review time by more than 20%, with 91% accuracy, while lawyers retain approval responsibility.
  • Osaic approved Nitrogen’s adviser tools, Masttro opened read-only AI connections to wealth data, and Glide plans to offer its portfolio-analysis tool from Nov. 30.
  • Analysts cite rising cloud revenue forecasts alongside weaker free-cash-flow estimates; Morgan Stanley expects hyperscaler spending of $1.6 trillion in 2028 and $1.8 trillion in 2029.
  • A University of Glasgow team backed by a £39 million UKRI initiative is developing AI methods to speed offshore-wind surveys and clarify uncertainty in subsurface predictions.
  • Anadyr Horizon CEO Arvid Bell says the company applies large-scale AI simulations to geopolitical and investment risk analysis.

SquadStack.ai says Maya beats leading AI models on sales calls

SquadStack.ai launched Maya, a large language model designed for sales conversations. The company says Maya scored 96.7% on its benchmark, ahead of Google’s Gemma 4 31B at 92.5%, DeepSeek V4.1 Flash at 92.2% and OpenAI’s GPT-6 Luna at 89%. SquadStack.ai trained Maya on more than 40 crore sales interactions and says it handles over 50 lakh conversations each day in Hinglish and other Indian languages. The company built its benchmark from 9,962 moments in 471 sales calls, using Anthropic’s Claude Haiku 5.5 to judge replies. Maya is built and hosted in India, and SquadStack.ai says it is developing ways to use live call cues such as pauses and changes in tone.

SquadStack.ai launches Maya, its sales model for live calls

SquadStack.ai introduced Maya, a large language model trained for live sales calls. The company says Maya scored 96.7% on its benchmark, which tested 11 models using 9,962 moments from 471 real calls. It was trained on more than 40 crore sales interactions collected over six years and handles over 50 lakh conversations a day in Hinglish and Indian languages. SquadStack.ai reports that Maya starts replies in a median of 278 milliseconds and is hosted in India, keeping customer conversations in the country. Maya runs within the company’s Voice AI Agents, and the company says its Lift system learns from live calls to improve replies.

SquadStack launches Maya for Indian voice sales agents

Noida-based SquadStack.ai launched Maya on 16 November 2026 as a language model for live sales calls. The company says Maya scored 96.7% on checks using 471 Indian sales calls, ahead of models from OpenAI, Google and DeepSeek. SquadStack says it trained Maya on more than 40 crore sales interactions collected over six years, and the model already handles more than 50 lakh conversations each day. Maya works inside SquadStack’s voice AI agents and is hosted in India, which the company says can help regulated businesses meet data rules. The company says Lift uses live call results to improve weak responses, and Maya is being offered in a closed beta to a small group of sales teams in India.

Two lesser-known AI infrastructure stocks with growth potential through 2030

The article highlights Jabil and Applied Digital as lesser-known companies that could benefit from growing demand for AI infrastructure. Jabil’s fiscal 2026 revenue rose 21% to $36 billion, and it expects revenue to reach $44.5 billion in fiscal 2027. It forecasts AI-related revenue of $22.1 billion in fiscal 2027, a 54% increase. Applied Digital’s fiscal 2027 first-quarter revenue rose 322% year over year to $342 million, and the company plans to deliver another 600 megawatts of data center capacity over the next 12 months. It aims to build 3.5 to 4 gigawatts of capacity by the end of 2030 and has a $36 billion contracted lease revenue pipeline tied to 1.4 gigawatts of capacity.

Jabil and Applied Digital draw attention as AI infrastructure stocks

The article points to Jabil and Applied Digital as AI infrastructure companies investors may want to consider alongside Nvidia and Micron. Nvidia and Micron supply chips used to support AI workloads in data centers, while Jabil and Applied Digital are presented as other potential beneficiaries of AI growth. The article says Jabil and Applied Digital could be worth holding through the end of the decade. The supplied text does not include further details about their financial results or forecasts.

AI video growth could drive emissions sharply higher by 2030

Researchers measured the yearly video output of two commercial text to video platforms and modeled the services’ environmental impact from training through storage. Their estimates suggest emissions could rise nearly 75 times by 2030 if video generation keeps growing, even as technology becomes more efficient. If use stays at 2026 levels, those improvements could cut emissions by about two fifths. The researchers call their estimates a conservative lower bound because they cover only one AI service among many.

AhnLab Cloudmate wins award for SecureBridge AI security tool

AhnLab announced on Nov. 16 that its subsidiary AhnLab Cloudmate won the Korea Startup Forum Chairman’s Award at the 21st Korea Internet Awards. The ceremony took place in Seoul on Nov. 13, and the award recognized SecureBridge, a tool that helps organizations use generative AI while protecting sensitive information. Its patented context analysis detects and blocks personal or confidential data in prompts, without requiring extra software agents. SecureBridge has standard level Software as a Service certification under South Korea’s Cloud Security Assurance Program and was named an outstanding information security technology for 2026 by the Ministry of Science and ICT and KISA.

Anadyr Horizon CEO discusses AI simulations for risk analysis

Arvid Bell, co-founder and CEO of Anadyr Horizon, discussed the company’s use of large scale AI simulation. The company is applying the technology to geopolitical and investment risk analysis.

Insurance companies hold $85.2 billion in AI debt as exposure grows

A financial entity created by Elon Musk and his advisers borrowed billions to buy AI chips and lease them to xAI, helping keep the debt off xAI’s official books. According to a Moody’s Ratings report, insurance company Athene bought $2.4 billion of the debt. U.S. life insurers held $85.2 billion in AI debt as of June 30, a 24 percent increase in six months, according to Moody’s. That was 2.1 percent of their total debt, but Moody’s says life insurers could hold up to $185 billion in AI investments by the end of next year.

Uber legal AI agent cuts contract review time by over 20 percent

Uber says its Legal Redlining Agent has cut average contract review time by more than 20 percent, with 91 percent accuracy in AI generated decisions. Built into Microsoft Word, it reviews proposed contract changes, checks legal risks and policies, and recommends whether lawyers should accept, reject, or change clauses. The system combines retrieval augmented generation, past lawyer feedback, and fixed rules, while lawyers remain responsible for reviewing and approving its recommendations. Uber began development in early 2024 and piloted the tool with its Legal team in 2025.

Osaic, Masttro and Glide Expand AI Tools for Wealth Management

Osaic approved Nitrogen’s full suite of adviser tools, including its Nucleus AI agent, for advisers across the Osaic network. The six products cover risk, research, income, taxes, legacy planning and insurance, and Nucleus can automate tasks such as analyzing tax returns. Masttro launched a model context Protocol server and developer portal that let eligible clients connect wealth data to third-party AI applications, with read-only access available from November 16. Glide launched a U.S. portfolio analysis tool that estimates asset allocation, returns and risk, and will make it available to investors and advisers starting November 30.

Analysts See AI Stocks Recovering as Data Center Debt Grows

Analysts say a recent sell-off in AI stocks reflects a repricing rather than a likely crash, but they disagree about risks from the huge spending on data centers. Generali’s Vladimir Oleinikov said revenue forecasts for major cloud companies rose by $235 billion for 2026 to 2028, while free cash flow estimates fell by $163 billion. Morgan Stanley called the sell-off an overreaction and still expects hyperscalers to spend $1.6 trillion in 2028 and $1.8 trillion in 2029, with AI returns on invested capital above 30 percent. TS Lombard analyst Dario Perkins warned that AI growth depends largely on spending within the sector rather than outside customers; the article also notes that about 20 percent of expected 2028 data center capacity is delayed until 2029.

University of Glasgow Researcher Develops AI for Faster Offshore Wind Surveys

Dr. Roisin Buckley of the University of Glasgow received funding through a £39 million UKRI initiative to develop AI tools for offshore wind seabed and subsurface surveys. Her team will combine AI with physics, geology and engineering to use geological, geophysical and engineering data, identify knowledge gaps and show how certain its predictions are. The aim is to cut the time and cost of investigations, which can take around 10 years and cost tens of millions of pounds. Working with Scottish Power Renewables, the team has assembled a large curated database of subsurface data, and the approach could also support transport, telecommunications, nuclear energy and carbon capture projects.

Sources

NOTE:

This news brief was generated using AI technology (including, but not limited to, Google Gemini API, Llama, Grok, and Mistral) from aggregated news articles, with minimal to no human editing/review. It is provided for informational purposes only and may contain inaccuracies or biases. This is not financial, investment, or professional advice. If you have any questions or concerns, please verify all information with the linked original articles in the Sources section below.

Artificial intelligence Voice AI Sales automation Large language models AI benchmarking Indian languages AI infrastructure Data centers AI semiconductors AI energy consumption Text-to-video generation AI emissions AI cybersecurity Prompt security AI financial risk Insurance investments Legal AI Retrieval-augmented generation Wealth management AI AI in renewable energy

Comments

Loading...