AI Furnace Newsroom: Google Brings Agents into Workspace, PepsiCo Hands Reins to AI, Amazon Joins Chip Wars

Welcome to this week’s AI Furnace Newsroom

This week in AI, we saw a culmination of the major trends of the year bubbling up as 2025 comes nearer to an end. This includes competition between foundation models increasing in velocity and pressure; hardware giants competing over chips as NVIDIA’s capacity continues to stay limited; job losses continue; startups move from AI pilots to internal scaling; enterprise CIOs focus on scale and figure out if they should build internally, partner with incumbents like Salesforce or bring in more agile startups to help them get ahead of competitors and target high ROI.

In today’s insights we cover:

1. Google Brings AI Agents Directly into the Workspace & Cracks Long-Term Memory

2. PepsiCo lets AI Make Decisions Without Human Approval

3. Amazon Makes Bold AI Chip Move

4. HP Cuts Jobs Due to Higher Chip Costs and Increased AI Adoption

5. IBM CEO Concerned About AI Data Centre Investments

Read time: 4 mins

💡 Furnace Insights

Google

Google Brings AI Agents Directly into the Workspace & Cracks Long-Term Memory

Google is on a roll, integrating agents into Workspace (Gmail, Drive, Docs and Chat) and separately solving AI’s long-term memory problem with the release of Titans in the same week. All this, hot on the heels of Google’s benchmark-beating Gemini 3 Pro and Nano Banana Pro launches.

Gemini 3 now drives customizable agents directly inside Gmail, Drive, Docs and Chat. Users can build multi-step agents in minutes with natural language prompts. Early testers report major time savings, with drafting cycles cut by up to 90%. People are starting to treat AI agents like extra team members, and Google is betting that putting them inside the tools people already live in will push that shift into the mainstream.

Google also launched Titan, an architecture that is said to give models real long-term memory during inference. Early tests show strong performance on long-context benchmarks, even beating larger models on tasks that require reading millions of tokens. Long-term memory has been the big missing piece for years. If these optimizations are sustained at scale, we may finally see AI that keeps stable context across long stretches instead of rebooting every session.

PepsiCo

PepsiCo lets AI Make Decisions Without Human Approval

Where teams previously needed meetings and human sign-off for AI actions, now AI agents at PepsiCo can execute those changes instantly, grounded in real-time data. This means faster replenishment, more accurate fulfillment, and supply-chain agility at global scale.

These deployments include:

• Demand forecasting that adjusts in real-time without waiting for weekly meetings

• Replenishment cycles that self-optimize based on velocity shifts

• Marketing budget allocation that moves spend automatically when performance drops

• Creative iteration that generates and tests variants without creative briefs

Amazon

Amazon Makes Bold AI Chip Move

In a surprise move, Amazon released its Trainium3 chip which is 4x more powerful than their previous model, and up to 50% cheaper compared to Nvidia’s GPUs. Amazon also released three new AI agents aimed at cutting down operational burden for AWS deployments, a trillion-dollar market.

The Trainium3 launch pushes Amazon into the same bracket as Google with its TPUs, as both pose a threat to Nvidia’s GPU dominance.

HP

HP Cuts Jobs Due to Higher Chip Costs and Increased AI Adoption

The PC maker HP announced it would cut 4,000-6,000 employees through 2028 as it plans to apply more usage of AI tools to areas like product development, customer support, and sales and marketing.

The company has recently benefited from increased demand for new AI PCs, but is also dealing with higher memory chip prices cutting into profits. Layoffs have also recently been announced at Apple, which cut some sales jobs, and at Verizon, with the telecoms giant cutting 13,000 jobs. A recent KPMG survey showed that 52% of U.S. workers fear AI will replace their jobs, a fear that’s nearly doubled from last year.

IBM

IBM CEO Concerned About AI Data Centre Investments

IBM CEO Arvind Krishna questioned the financial assumptions around AI data centre investments and broke down the economics.

He estimated that a 1GW site costs around $80B to populate with hardware. With current plans to build 100GW worth of capacity, that would require $8T of investment. On top of this, most of the high-end GPU hardware deployed in these centres depreciates over roughly 5 years, at which point operators need to replace them in full.

Krishna calculated that servicing the cost of capital for these multi-gigawatt campuses would require hundreds of billions of dollars in annual profit just to remain neutral. The question being asked is whether AI will ever generate enough economic value to justify an $8T rebuild every 5 years.

Flock’s AI helps fashion brands go from manufacturing to marketing instantly through customized models and apparel consistency

Flock provides fashion and beauty brands with an AI platform to create customized imagery that reflects customer and brand aesthetic. Generate production-quality models, imagery and apparel layering, with never-before-seen speed and brand consistency for the fashion and beauty industry.

Why it matters: It’s faster, easier and higher converting than a manual photoshoot, providing 10x cost savings and 30%+ average conversion increase.

Check out Flock AI.

📈 AI Venture Deals of the Week

• Artera, a AI-powered patient communications company, raised $65M growth funding.

• Confluent, a real-time data streaming platform, in talks with IBM to be acquired for $11B amid AI push.

• imper.ai, a real-time cyber impersonation prevention company, raised $28M funding.

• Gradium, a AI voice startup, raised $70M Seed funding.

• Aaru, a synthetic-research AI startup, raised a Series A at a $1B headline valuation.

• Scowtt, an AI-powered advertising optimization company, raised $12M Series A funding.

• Limitless, an AI-wearable device startup, was acquired by Meta.

⚒️ New AI Product Launches You Don’t Want to Miss 

• Amazon Nova Forge by Amazon Web Services, a service for building custom AI models using your own data.

• Poetiq ARC‑AGI by Poetiq, a meta-system combining multiple LLMs to solve complex reasoning tasks cost-effectively.

• VibeVoice by Microsoft, a framework for generating long-form, multi-speaker conversational audio (e.g. podcasts) from text.

• Colab Data Explorer, a tool to search Kaggle datasets, models and competitions directly within Google Colab.

• Conductor by Orkes, a tool to run multiple code-generating agents in parallel and manage their output/workspaces.

Upcoming Events 📅

• AIF Founder Supper Club @ Sicily Osteria - Tuesday, December 9

• AIF Founder Supper Club @ Whiskey Club - Wednesday, January 21

• AIF Founder Supper Club @ Wine Bar - Wednesday, February 18

Interested in meeting the Who’s Who of AI?

The AI Hot 100 Summit is back for its third edition on May 7th, 2026 in New York City. The only AI Summit built for real connections where startup innovation meets enterprise implementation. Join over 500 AI Executive Leaders, Founders and Investors for a jam packed day of enterprise AI lightning talks, innovation showcases, and curated networking.

Get ready to meet and greet speakers and attendees from OpenAI, NVIDIA, Haleon plc, Legora, Anthropic, Cursor, Eleven Labs, Adobe, Walmart, Insight Partners & more.

🗓️ May 7, 2026

📍New York City