Altman Rejects Musk’s $97.4B Offer for OpenAI; Counteroffers for X Instead

Altman Rejects Musk’s $97.4B Offer for OpenAI; Counteroffers for X Instead

Elon Musk desperately wants to buy OpenAI. Well, at least it’s a non-profit division. And he’s willing to throw $97.4 billion at it.

According to The Wall Street Journal, Musk and a group of investors just made a formal bid. The move? Bold. The timing? Messy. The drama? Unmatched.

And Sam Altman? Not interested. At all.

The Offer, The Shade, and The Chaos

Musk’s lawyer, Marc Toberoff, submitted the offer to OpenAI’s board. Not long after, Musk himself made his stance clear:

"It’s time for OpenAI to return to the open-source, safety-focused force for good it once was."

His plan? If the deal goes through, merge OpenAI with xAI, his own AI company. Investors backing him include Valor Equity Partners, Baron Capital, Atreides Management, Vy Capital, 8VC, Ari Emanuel, and Palantir’s Joe Lonsdale.

Altman, though? He clapped back on Musk’s own platform, X (formerly Twitter):

OpenAI vs. Musk: A Rivalry With History

Musk and OpenAI go way back. He helped found OpenAI in 2015. Left in 2018. Didn't like the direction it was going.

By 2019, OpenAI ditched its non-profit status to take in big investors. Microsoft jumped in. Big money started flowing. The company shifted. Musk hated it.

Now? He’s throwing lawsuits at them. Says OpenAI and Microsoft are trying to monopolize AI. Claims they’re blocking investors from funding competitors like xAI.

The Altman Firing Saga (Yeah, That Happened Too)

Oh, and let’s not forget that time OpenAI fired Altman.

November 17, 2023—Boom. He was out. The board said he wasn’t “consistently candid.”

Five days later? OpenAI reversed course. Brought Altman back. Revamped the board that fired him.

The takeaway? OpenAI is a battlefield.

So What Now?

Musk’s not getting OpenAI. Not without a fight.

Altman’s doubling down. Says OpenAI’s structure ensures nobody can take control. Calls Musk’s bid a distraction. A tactic to slow them down. One thing’s for sure—this fight is far from over.

Think Before You Prompt: 10% of Employee AI Queries Contain Private Data

Employees are leaking sensitive data. Not just through shady, unauthorized AI apps. But even through the ones IT has approved.

A new report from Harmonic breaks it down. 8.5% of AI queries include sensitive info. That’s bad. Really bad.

What’s getting leaked?

The usual suspects.

This isn’t just a privacy issue. It’s a legal time bomb.

Shadow AI: The CISO’s Worst Nightmare

AI use in companies falls into three messy categories:

1. Sanctioned AI – IT-approved, enterprise-grade AI tools.

2. Shadow AI – Free, unauthorized AI apps. The wild west.

3. Semi-shadow AI – Paid tools that management sneaks in without IT approval.

Shadow AI is a problem. Semi-shadow AI? Maybe worse. Employees think it's legit. But it’s not.

Free AI tools? The worst offenders. 54% of leaks came from ChatGPT’s free tier.

But Paid AI is Safe, Right? Nope.

Even “trusted” AI tools aren’t foolproof. Sure, they say they won’t train on user data. But legal experts aren’t buying it.

Take trade secrets. Once an employee enters one into AI? Legal protection could be gone. If a competitor finds out? They can argue it’s public knowledge.

And guess what? A contract promising data protection isn’t enough. Enterprises need real safeguards.

Why Are Employees Using Shadow AI?

Simple. IT isn’t giving them what they need.

CISOs can’t just crack down. They need to fix the root issue.

The Old Playbook Is Failing

Traditional monitor-and-block security? Useless. AI adoption outpaced IT. Employees are moving faster than security teams can catch up.

Kaz Hassan from Unily puts it bluntly:

Data Leaks Work Both Ways

It’s not just sensitive info getting out. It’s also bad AI data coming in.

Hallucinated reports. Misinformation. AI-generated errors seeping into corporate decisions.

CISOs shouldn’t just worry about what’s leaving. They need to worry about what’s being brought in, too.

The AI revolution is here. Companies better catch up.

Danger Bells For NVIDIA? OpenAI’s First AI Chip is Coming Soon

OpenAI is going all in. They’re making their own AI chip. No more full reliance on Nvidia.

According to Reuters, the design will be finalized in the next few months. Then? It’s off to TSMC for fabrication.

The Process Ain’t Easy

First step: “Taping out”—tech jargon for sending the final chip design to be manufactured. Sounds simple. It’s not.

It takes six months and costs tens of millions. And here’s the kicker—no guarantee it works. If something’s off? OpenAI has to debug, redesign, and do it all over again.

What’s The Plan?

If all goes well, mass production could start by 2026.

The chip will be built using TSMC’s 3-nanometer process. It’ll use a systolic array architecture with high-bandwidth memory—the same tech Nvidia swears by.

But OpenAI isn’t going all-in just yet. The chip will be mostly used to run AI models, not train them. A test run before bigger things.

Why Bother Making Their Own Chip?

AI chatbots burn through chips like crazy. Training massive AI models takes a ton of computing power.

Right now, Nvidia dominates the market—around 80% share. That means big AI companies have little bargaining power. OpenAI? They’re playing the long game.

If their first chip works? Expect more. Bigger. Better. Faster.

Is OpenAI Beating Big Tech?

It sure looks like it.

Meta and Microsoft? They’ve tried. And failed. Years of effort. No breakthrough.

Meanwhile, OpenAI is moving fast. And they’re already ahead of the pack.

But here’s an interesting twist—DeepSeek is proving you don’t even need crazy chip power to build strong AI. If that trend holds? The whole game could change.

For now, though? All eyes are on OpenAI.

Alibaba’s Qwen Models Are Running the Show And Open-Source AI Knows It

Alibaba’s AI game is stronger than ever.

According to Hugging Face, the top 10 open-source LLMs are all powered by Qwen models. That’s every single one.

The latest Open LLM Leaderboard? Dominated by Alibaba’s Qwen2.5-72b series. Even the #1 model, calme-3.2-instruct-78b, is just a fine-tuned version of Qwen.

What’s the Big Deal?

LLMs are the backbone of Gen AI—the tech behind chatbots, AI-generated content, and everything in between.

Alibaba Cloud’s open-source LLMs have fueled over 90,000 models on Hugging Face. That’s a massive influence.

And China? It’s flexing hard in AI.

Alibaba vs. The Rest

Just last September, Alibaba dropped Qwen2.5, releasing over 100 open-source models with sizes ranging from 500M to 72B parameters. That’s serious range.

At launch, Qwen2.5-72b outperformed Meta’s Llama3.1-405b in benchmarks.

Meanwhile, DeepSeek? They’ve been making waves too. Their AI models, DeepSeek-V3 and DeepSeek-R1, got attention for costing way less to train. But Alibaba? Still leading.

Even the “Godmother of AI” Is on Board

Li Feifei, one of AI’s biggest names, just trained a reasoning model for under $50—using Alibaba’s Qwen2.5-32b-Instruct.

And Alibaba’s closed-source model? The Qwen2.5-Max? Already ranked #7 on Chatbot Arena. DeepSeek-V3? Still stuck at #9.

Stocks and Market Moves

Alibaba’s stock? Steady in Hong Kong at HK$104.90. But in New York? Up 7.6% overnight.

What’s Next?

Alibaba’s dominating open-source AI. China’s AI scene? Only getting deeper. And with Qwen running the show, big tech better pays attention.

PM Modi Pushes for Open-Source AI & Fair Data at Paris AI Summit

PM Narendra Modi took the stage at the AI Action Summit in Paris. His message? AI needs to be open, unbiased, and sustainable.

“We need trust and transparency in AI,” he said. Open-source systems. Clean, bias-free datasets. Tech for the people. Not just the powerful.

India’s Taking the Lead

India’s not just talking. They’re co-chairing the AI governance theme with France and Canada. After the UK’s AI Safety Summit (2023) and Korea’s AI Seoul Summit (2024), this is India’s big global AI move.

Modi also flagged a big problem—AI eats up massive energy. The solution? Green power. But that’s not enough. AI models also need to be leaner, faster, and smarter.

AI vs. Jobs? History Says Relax.

People fear AI will kill jobs. Modi says, not so fast.

Tech doesn’t erase work—it changes it. New roles emerge. Old ones evolve. The real need? Reskilling. AI-driven jobs are coming. India’s getting ready.

India’s Building Its Own AI Model

Big news. India’s making its own LLM.

Announced on January 30, India’s Gen AI project will compete with ChatGPT, DeepSeek, and others. 18,693 GPUs are locked in. The India AI Compute Facility will fuel it.

Public-private partnerships? Also in play. Pooling resources. Sharing compute power. A whole AI ecosystem is growing.

AI Must Be Rooted in Local Context

Modi pushed data localization. Tech must fit local needs. Global AI won’t work everywhere.

He also pointed out a wild fact—the human brain writes poetry and builds spaceships on less energy than a light bulb. AI? Nowhere close.

AI Future? Humans Hold the Key.

Some worry machines will outthink us. Modi doesn’t buy it.

India’s in the AI game. And it’s playing to win.

Thank you for reading

-Shen & Towards AGI team