It Feels Like a New AI Story Drops Every Day…
AI is evolving so fast, it’s like every other day there’s a groundbreaking update. Startups are scrambling to stand out, and competition—especially in China—is heating up at lightning speed.
But let’s set the scene first.
The Chinese AI Race: DeepSeek Sparks a Revolution
Earlier this year, Chinese AI company DeepSeek released a super cost-effective AI model that shook up the entire market. Suddenly, the game changed. Startups began launching AI products at full throttle. Local governments started throwing in millions in funding. Everyone wanted to become China’s top AI player.
One name that’s been making headlines again and again is Zepu AI. Though they’re not brand new (they spun out of a Tsinghua University lab in 2019), they’ve been ramping up fast, releasing several models under their GLM series. Now, they claim their latest model, GLM4, outperforms GPT-4 in multiple benchmarks. That’s a huge deal.
AutoGM Rumination: Zepu’s Free AI Agent
Zepu just launched a free AI agent called AutoGM Rumination. It’s like your all-in-one assistant: research tasks, web browsing, trip planning, writing—whatever you need.
And here's the kicker—it’s completely free.
Compare that to their competitor, Manis, which charges up to $199/month for similar services. In a market that’s become super cutthroat, offering a powerful free tool is a bold, disruptive move.
Homegrown Tech That’s 8x Faster Than DeepSeek
Zepu’s CEO, Jiang Peng, said at a recent event in Beijing that AutoGM Rumination runs on their own models, including GLM Z1 Air—a system they claim matches DeepSeek’s top model Star R1 in performance but runs eight times faster and uses far less computing power.
That’s not just a brag—it’s a big deal. Efficiency matters when it comes to large-scale AI deployment.
Heavy Government Support, But Major Challenges Too
Zepu’s success has a lot to do with strong financial support from local governments. Just recently, the city of Chengdu invested 300 million yuan (~$41.5 million) in Zepu—their third round of funding in a single month. They’ve also raised 1 billion yuan in another round, plus 500 million yuan from the Huafa Group, a state-owned enterprise.
But it hasn’t all been smooth sailing.
In early 2024, Zepu was added to the U.S. export control list, meaning they can’t buy critical U.S.-made tech—especially chips from companies like NVIDIA. That’s a huge obstacle for any AI company.
How Other Chinese Startups Are Reacting
Zepu isn’t the only one trying to adapt. The Chinese AI scene is shifting fast:
-
01.AI, run by former Google China head Kai-Fu Lee, stopped training its own models and now offers services based on DeepSeek’s open-source tech.
-
Bichuan fired its sales team focused on banking and pivoted to healthcare AI.
-
Moonshot, known for its chatbot Kimi, faced reliability issues and is refocusing on model training.
These shifts show just how much DeepSeek’s approach has influenced the entire ecosystem. In fact, some say the Chinese government now favors DeepSeek for official projects.
Zepu’s High-Risk, High-Reward Play
Zepu hasn’t backed off. They’re still releasing consumer apps, building enterprise solutions, and reportedly preparing for an IPO on the Shanghai Star Market.
Here’s what we know:
-
In 2024, Zepu made around 300 million yuan in sales.
-
But they also lost about 2 billion yuan—a massive gap.
-
They have 800 employees, far more than DeepSeek’s 160.
That kind of spending, even with government backing, makes investors nervous. Especially now, with DeepSeek proving you can succeed with a leaner, more efficient model.
Will Zepu’s Business Model Survive?
A lot rides on whether Zepu can maintain its momentum. One big question: If local governments start favoring DeepSeek for AI solutions, what happens to Zepu’s B2G (business-to-government) revenue stream?
They’ve built a big part of their business on selling AI services to local authorities. If those deals dry up, it could threaten their entire growth strategy.
Meanwhile, in the U.S…
AI developments aren’t limited to China. In the U.S., Apple is reportedly working on an AI health assistant called Project Malberry, building on its earlier Project Quartz.
It’s expected to launch with iOS 19.4 sometime in mid-2025. Malberry could track health data from iPhones and Apple Watches to offer custom fitness plans, diets, sleep guidance, and even food tracking. That puts Apple in direct competition with fitness apps like MyFitnessPal.
They’re even training the AI with help from real doctors.
Of course, this raises privacy concerns—but if done right, Apple could take a giant leap into AI-powered healthcare.
Elon Musk’s $52 Billion AI Power Move
In another jaw-dropping shift, Elon Musk just sold X (formerly Twitter) to his AI company xAI in a $52 billion all-stock deal.
Why? Integration.
With 600 million users on X and a cutting-edge AI lab under one roof, Musk wants to embed AI into the social media experience—recommendations, content moderation, smart replies, and more.
He says it's about “seeking truth” and “advancing knowledge,” but whatever the motive, this deal shows just how far AI is reaching into every digital space.
So, What’s the Bottom Line?
AI isn’t just a buzzword anymore. It’s reshaping business models, industries, and even national tech strategies.
We’re seeing:
-
More free AI agents
-
Bigger government-backed fundings
-
Startups pivoting or doubling down
-
Companies like Apple and Musk’s xAI redefining their core products through AI
And as AI evolves, the need for speed, efficiency, privacy, and strategy will only grow.