☄️Inside the AI hardware shift: Altman U-turns, Baidu goes open, and Google shrinks the cloud

Sam Altman Says Today’s Computers Can’t Handle Tomorrow’s AI.

OpenAI is pouring billions into new hardware efforts, teaming with Jony Ive and designing a mysterious screenless device. Meanwhile, a Chinese tech giant makes a bold open-source move, a major new AI model hits your phone, and one CEO says machines already do half the work.

📌 In today’s Generative AI Newsletter:

• Altman reverses course, says new hardware is needed for AGI

• Baidu open-sources Ernie, challenging Western AI giants

• Google’s Gemma 3n brings real multimodal AI to your phone

• Salesforce says AI now does 30–50% of its internal work

Special highlight from our network

The #LeadWithAIAgents Hackathon isn’t just about learning. It's about being seen.

✅ $10,000 in prizes

✅ Badges & certificates for your portfolio

✅ Global visibility across 100+ countries

✅ Projects that stand out in hiring & funding

Real tools, real teams, real outcomes.

Let the world see what you can do with AI agents & orchestration.

📍Online

🔗 Join now as a builder, an ambassador or a mentor 

Special highlight from our network

AI Trust Hub: Get Future-ready By Mastering AI Risks

Whether you’re a CIO, security architect, or aspiring AI risk leader, Tumeryk’s AI Trust Hub offers tailored, self-paced courses to tackle the real-world challenges of GenAI risk management.

Learn to:

• Navigate AI risk, governance, and compliance frameworks
• Implement ISO 42001, the EU AI Act, and NIST 600 standards
• Understand AI vulnerabilities, risk scoring, and governance best practices
• Build practical expertise for career growth in an AI-driven world

Designed for both executives and hands-on practitioners.

🚀 Launch offer: Use code GENAI2025 for 50% off — limited time only.

👉 Learn more

🖥️ Sam Altman U-Turns on AI Hardware, Says Current Computers Can’t Keep Up

After years of insisting AI wouldn’t need new hardware, Sam Altman is changing his tune.

The OpenAI CEO now says the computers we rely on today are fundamentally unfit for the AI era and he’s backing that belief with billions.

Speaking on a podcast with his brother Jack, Altman admitted that modern machines were “designed for a world without AI,” and that the shift now demands “a different kind of hardware.” 

That’s a sharp pivot from his earlier stance that software alone could drive the rise of AGI.

OpenAI is already acting on the new vision. It’s building custom chips and co-developing a screen-free AI device with ex-Apple design chief Jony Ive. 

The mystery product won’t be a phone or wearable, but a contextual AI “companion” built from the ground up to rethink human-machine interaction.

OpenAI recently spent $6.5B to acquire io Products, a company co-founded by Ive and focused on AI-native hardware.

The era of general-purpose devices is fading. The future of AI won’t run on yesterday’s machines.

🚨Baidu Open-Sources Ernie AI: China's Next DeepSeek Moment?

Baidu is releasing its flagship AI model Ernie to the open-source community starting today, in a move that could mark China’s second “DeepSeek moment.” 

The release comes just months after the company launched Ernie 4.5 and its first reasoning model, Ernie X1. 

While details remain unclear on whether the release will be fully open-source or limited to model weights, it still marks the first time a Chinese tech giant has opened its core model to the public.

If the release includes full model weights, training data, and architecture, Baidu could radically shift the balance in AI.

Last time this happened with DeepSeek-R1, U.S. AI stocks took a hit, and global competition surged.

Now, with Ernie potentially offering comparable performance and backed by a massive user base in China, it could push Western players to reconsider their closed-garden strategies.

Baidu is using openness as leverage and a calculated threat to OpenAI, Anthropic, and Google: keep your walls up, and watch the rest of the world route around you.

📱 Google’s Gemma 3n Shreds the Cloud. AI Is Now Truly On-Device

Google just fired a major shot in the on-device AI wars with Gemma 3n, a lightweight, open-weight multimodal model that runs right on your phone, laptop, or tablet, no internet, no middleman, no strings. 

Google’s Gemma 3n is already live, downloadable, and translating speech, analyzing video, and parsing images in real time on consumer hardware.

Gemma 3n comes in 5B and 8B parameter versions and can run on just 2 to 3GB of RAM, thanks to Google’s custom architecture (MatFormer, MobileNet-V5, and Per-Layer Embeddings). 

It handles audio, video, images, and text natively, with translation across 35 languages, real-time scene recognition at 60FPS, and accessibility tools that work offline. 

The larger model even crossed the 1300 mark on LMArena, a rare feat for sub-10B models.

While OpenAI talks about local AI, Google is already shipping it. Gemma 3n is now available via Hugging Face, Kaggle, and Google AI Studio, putting pressure on every closed-shop lab still charging for basic features. 

AI that lives in your pocket just got a whole lot smarter and freer.

💼 AI Does Half the Work at Salesforce. But Don’t Call It Mass Unemployment Yet

Salesforce CEO Marc Benioff says AI is now doing 30 to 50% of the company’s work, from software engineering to customer support. 

With AI driving real efficiency gains and cutting back hiring needs, Benioff dubbed it a “digital labour revolution” and told Bloomberg it’s time for workers to "get their heads around" the idea that machines are taking over more of the job.

The company’s AI tools are already clocking 93% accuracy in customer service tasks, deployed by big-name clients like Disney. 

But even as bots take the wheel, Salesforce is redirecting human effort, cutting technical hires while planning a 22% increase in sales roles. 

What’s unfolding mirrors every major labor shift of the past century. Just like tractors didn't kill farming entirely.

 McKinsey data backs this up: while half of tasks could be automated, less than 5% of jobs are fully replaceable with today’s tech.

What’s emerging instead is a workforce divided by role and readiness. Physical, rule-based jobs are getting offloaded. 

Still sticky. Benioff’s numbers are real, but for most companies still crawling into AI, the revolution is on the runway not yet in the skies.