The Chinese AI large model sector is witnessing a stark divergence in fortunes, creating a compelling pairs trading opportunity. Since the end of March, shares of Knowledge Atlas (ASX: 02513) have surged approximately 170%, while those of MiniMax have fallen by about 50% over the same period. This dramatic split in performance has led to a highly attractive market strategy: going long on Knowledge Atlas and shorting MiniMax.

Core Drivers of Divergence

The fundamental driver of this divergence lies in a stark contrast in profit expectations. Knowledge Atlas has managed to increase prices for its GLM series models this year while maintaining sales volume. Consequently, the market consensus for its sales expectations has skyrocketed nearly 250% since the start of the year, now surpassing that of MiniMax. In contrast, MiniMax announced a 50% price cut for its flagship M3 model just one week after its launch, a move that prompted major investment banks like Goldman Sachs and JPMorgan to downgrade their ratings or price targets for the company.

Upcoming Unlock Event

An impending wave of IPO lock-up expiries in early July could further amplify the momentum for this pairs trade. According to estimates from HSBC Holdings, approximately 65% of MiniMax's total share capital will become freely tradable on July 8th. In comparison, Knowledge Atlas will see only about 6% of its shares unlocked the previous day. Analysts note that the larger proportion of shares becoming available will make MiniMax stock easier to borrow, thereby lowering the cost to short it and giving bearish positions a greater advantage.

Recent US government restrictions on foreign access to top-tier products from companies like Anthropic have redirected more investor attention to the Chinese AI sector. Knowledge Atlas capitalized on this momentum by releasing its most powerful open-source model to date, GLM-5.2 (Max), propelling its share price to a weekly gain of over 70% and widening the performance gap with MiniMax.

Sharp Split in Performance and Expectations

Both Knowledge Atlas and MiniMax listed on the Hong Kong stock exchange in January. Their shares initially moved higher in tandem, ranking among the top three performers in the Hang Seng Tech Index this year. However, subsequent model updates and pricing adjustments have led to a significant divergence in their trajectories.

Year-to-date, Knowledge Atlas has surged more than 1500%, making it the standout performer among Hang Seng Tech Index constituents. Conversely, MiniMax has declined roughly 50% since late March. Leonid Mironov, a portfolio manager at Gavekal Capital Ltd., described the long-Knowledge Atlas/short-MiniMax trade as "a very profitable trade," noting that Knowledge Atlas's outperformance is unsurprising as it "has a better-quality model and structurally better profit potential."

The disparity is equally stark in sales expectations. Bloomberg data shows that the market consensus for Knowledge Atlas's sales has surged nearly 250% this year, now exceeding the corresponding expectation for MiniMax, indicating a fundamental divergence in the market's assessment of the two companies' commercial prospects.

Pricing Strategies Prompt Analyst Actions

The two companies have adopted diametrically opposite pricing strategies. Knowledge Atlas raised prices for its GLM series models this year while maintaining sales scale, demonstrating strong product pricing power. In contrast, MiniMax announced a 50% price cut just one week after launching its latest flagship M3 model, aiming to attract more users.

This pricing action triggered a swift reaction from Wall Street institutions. Goldman Sachs, in a research report this week, lowered its price target for MiniMax by 14%, citing the negative impact of price cuts on margins. JPMorgan analyst Olivia Xu raised the target price for Knowledge Atlas while downgrading her view on MiniMax, characterizing its lower pricing as "a signal of lower-than-expected model capabilities." A MiniMax spokesperson did not respond to a request for comment on these market views.

Meanwhile, model performance data supports Knowledge Atlas's pricing confidence. According to benchmark platform Arena.AI, the GLM-5.2 (Max) model from Knowledge Atlas ranks second globally in front-end code writing capabilities, just behind Claude Fable 5. Mironov added that Knowledge Atlas is "better at convincing the market that its customers are stickier."

Unlock Pressure as a Key Catalyst

The expiration of IPO cornerstone investor lock-up periods in early July is viewed by analysts as a key node for the further development of this pairs trade. The stark difference in the scale of share unlocks will directly impact the liquidity profile and short-selling costs for both stocks. A large influx of new shares into the market will make MiniMax stock easier to borrow, lowering borrowing costs and providing a cheaper window for short sellers. Reports indicate the Hong Kong market faces lock-up expiry pressure of up to $33 billion, with the effect being particularly concentrated in the AI sector.

Risks and Counterpoints

Despite shorting MiniMax becoming a popular market play, some institutions have cautioned against overly one-sided bets. Bank of America Securities this week initiated coverage on both Knowledge Atlas and MiniMax with Buy ratings, suggesting MiniMax could catch up after digesting the unlock pressure.

Felix Wang, head of the technology sector at Hedgeye Risk Management, acknowledged that "shorting MiniMax is quite popular right now" but emphasized that both companies face similar long-term challenges. He is the only analyst tracked who holds an equivalent Sell rating on Knowledge Atlas. His key arguments include competitive threats from DeepSeek's lower pricing, with its V4 model launch reportedly triggering a price war in China's AI industry that tests Knowledge Atlas's pricing power. He also noted that subsequent IPO supply pressure cannot be ignored, as domestic competitors like Kimi Chat from Moonshot AI and major US AI companies have potential listing plans that could dilute capital attraction for existing stocks.