DeepSeek's CEO has been absent from public view for more than a year, yet the Chinese AI startup continues to dominate global markets with increasingly aggressive pricing and powerful new models. Founder and CEO Liang Wenfeng last appeared publicly in February 2025 during a televised meeting with Chinese President Xi Jinping, immediately after releasing the company's R1 reasoning model. Since then, silence.

The absence has created an unusual leadership vacuum for a company that just launched its V4 model in April 2026, marking a significant milestone in its rapid ascent. While Liang's control over DeepSeek appears unchanged, recent corporate filings and social media activity suggest the company is adjusting how it communicates with the world.

Who Is Speaking for DeepSeek Now?

Senior researcher Chen Deli has emerged as DeepSeek's new public face, posting to his nearly 24,000 followers on X (formerly Twitter) following the V4 release. Chen wrote, "484 days later, we humbly share our labour of love. As always, we stay true to long-termism and open source for all," signaling the company's continued commitment to accessibility despite its explosive growth.

Chen

For a company known for operating in the shadows, this shift represents a notable change in strategy. DeepSeek has historically maintained a low profile compared to Western AI labs, but the emergence of Chen as a spokesperson suggests the startup is preparing for increased public scrutiny as it competes directly with OpenAI, Anthropic, and Google.

What Do DeepSeek's Corporate Changes Reveal About Its Future?

Despite Liang's absence from public life, his grip on the company has actually strengthened. According to corporate database Qichacha, Liang's equity stake recently increased dramatically to 34 percent from just 1 percent. His subscribed capital jumped from 100,000 yuan (approximately $14,630 USD) to 5.1 million yuan, while DeepSeek's registered capital rose from 10 million yuan to 15 million yuan.

These structural changes suggest that Liang remains firmly in control of strategic decisions, even if he's delegating public-facing responsibilities to others like Chen. The timing of these equity adjustments coincides with DeepSeek's most aggressive market push yet, raising questions about whether the CEO is deliberately stepping back from the spotlight to focus on long-term strategy.

How DeepSeek Is Reshaping the AI Market Through Pricing

While leadership questions swirl, DeepSeek's business strategy speaks louder than any public statement. The company has positioned itself as the most affordable serious AI lab in the world, with pricing that undercuts competitors by staggering margins.

• V4 Flash Pricing: The flagship model costs $0.14 per million input tokens and $0.28 per million output tokens, making it roughly 35 to 100 times cheaper than GPT-5.5 or Claude Opus 4.7
• V4 Pro Launch Promotion: The company is offering a 75 percent discount on its reasoning model until May 5, 2026, at $0.435 per million input tokens and $0.87 per million output tokens
• Free Tier Access: DeepSeek's web chat at chat.deepseek.com remains completely free for individual users with no paywall, Plus plan, or Pro tier restrictions
• Developer Incentives: New API users receive 5 million free tokens valid for 30 days, worth roughly $8.40 at current V4 Flash rates, with no credit card required
• Context Window: Both V4 models support a 1 million token context window at no extra charge, allowing users to process roughly 750,000 English words per request

This aggressive pricing strategy directly challenges the profitability models of Western AI companies. DeepSeek's stated goal is rock-bottom prices, backed by custom Huawei Ascend and Cambricon chips that reduce infrastructure costs compared to Nvidia's dominant GPUs.

Steps to Optimize Your DeepSeek API Costs

• Pin Your System Prompt: Cache hits cost 98 percent less on V4 Flash, so keep the same opening context across multiple requests to maximize savings
• Choose V4 Flash by Default: Use the cheaper Flash model for most tasks and only escalate to V4 Pro when you need deeper reasoning capabilities
• Leverage Off-Peak Discounts: DeepSeek historically offers 50 to 75 percent additional discounts during off-peak hours from 16:30 to 00:30 UTC, though availability may vary for V4
• Use Prompt Caching Strategically: Cache hits can reduce bills by 80 percent or more when you're processing similar content repeatedly

For high-volume workloads like content pipelines, summarization, classification, and retrieval-augmented generation (RAG) over large document collections, DeepSeek is often the only model that makes the unit economics work at scale.

Why Leadership Transparency Matters in AI

The mystery surrounding Liang's whereabouts raises broader questions about governance and accountability in the AI industry. Unlike OpenAI's Sam Altman or Anthropic's Dario Amodei, who maintain visible public profiles, DeepSeek's founder has chosen near-total anonymity while his company reshapes the global AI market.

This opacity contrasts sharply with DeepSeek's commitment to open-source development and transparent pricing. The company has released its R1 reasoning model and other systems to the public, yet keeps its own leadership structure largely hidden from view. Whether this reflects cultural preferences, regulatory caution, or strategic calculation remains unclear.

As DeepSeek continues to challenge Western AI dominance with superior pricing and competitive performance, the emergence of Chen Deli as a public spokesperson may signal a shift toward greater transparency. However, until Liang reappears or the company provides clarity on his role, questions about DeepSeek's governance will likely persist alongside its rapid market expansion.