OpenAI’s Deepening Deficit and Faltering Growth: A Push for ‘Subscriber Lock-In’ Through Low-Cost Plans

Expansion of Low-Cost Plans in Emerging Markets
$11 Billion in Accumulated Losses
User Attrition Sparks Defensive Moves

OpenAI has launched a new low-cost subscription plan priced at $5 per month in emerging markets—an experiment aimed ostensibly at widening accessibility but, in essence, viewed as an urgent bid to shore up cash flow amid mounting losses and decelerating growth.

Targeting Emerging Markets with Untapped Subscriber Potential

According to U.S. tech outlet TechCrunch on the 19th (local time), OpenAI’s budget plan “ChatGPT Go” has been rolled out to 16 additional countries across Asia, including Thailand, Vietnam, Bangladesh, Cambodia, Laos, Malaysia, Nepal, Pakistan, and the Philippines. In select markets such as Malaysia, Thailand, Vietnam, the Philippines, and Pakistan, users can pay in local currency, while others are billed directly in U.S. dollars at roughly the same $5 rate.

In a blog post, OpenAI described ChatGPT Go as “an affordable plan providing access to core ChatGPT capabilities” and noted that the offering would “gradually expand to all countries.” Compared with the free version, ChatGPT Go provides higher limits on response processing, image generation, file uploads, and data analysis, along with twice the memory capacity.

Analysts interpret the move as a calculated effort to penetrate high-potential emerging markets where lower income levels have constrained paid subscriptions. By introducing a budget-tier plan, OpenAI seeks to lower entry barriers, accelerate user monetization, and strengthen platform lock-in amid intensifying global AI competition. Following the plan’s launch, user growth has rebounded—India’s surge in active users helped lift ChatGPT’s weekly active user count from 700 million in August to 800 million last month.

OpenAI’s Deficit Nears $11 Billion

Despite a near-$500 billion valuation, OpenAI continues to bleed capital due to massive AI infrastructure investments. The company generated $4.3 billion in revenue during the first half of the year, surpassing its full-year 2024 figure by 16%, but spent $7.8 billion on R&D and operating expenses, resulting in substantial losses.

That includes roughly $2.5 billion in payments for Nvidia server rentals—double the previous year’s expenditure. Another $2.5 billion was issued as stock-based compensation, also double year-over-year, underscoring the fierce competition for AI talent.

Projections indicate OpenAI will post about $13 billion in annual revenue this year against $8.5 billion in cash burn. While it targets $200 billion in annual sales by 2030, cumulative server expenses alone are expected to reach $450 billion by then. To sustain operations, OpenAI has pursued successive multibillion-dollar fundraising rounds, including up to $100 billion from Nvidia—an amount still deemed insufficient given its cash drain.

Growth Plateau and Declining Engagement

Signs of stagnation are emerging. Data from analytics firm Apptopia show that global monthly downloads of the ChatGPT mobile app are slowing sharply, with October’s growth rate projected to fall 8.1% from the previous month. While total installs remain in the millions, momentum has clearly weakened. In the U.S., average daily usage time per active user has dropped 22.5% since July, while the average number of sessions per user fell 20.7%. Analysts attribute the slowdown to the rapid ascent of rivals like Google’s Gemini and controversies surrounding the GPT-5 model.

OpenAI attempted to reverse the trend with GPT-5—but instead found itself embroiled in backlash over inflated claims. On the 19th, OpenAI VP Kevin Weil announced on social media that GPT-5 had “solved ten long-standing Erdős problems,” but mathematician Thomas Bloom dismissed the claim as “a dramatic misunderstanding.” Bloom clarified that his website had listed those problems as “open” simply because he had not reviewed the relevant papers, stating, “GPT-5 didn’t discover new proofs—it merely retrieved existing ones.” The Erdős problems, introduced by Hungarian mathematician Paul Erdős, are notoriously difficult unsolved puzzles in combinatorics and number theory.

The controversy spread swiftly. Meta’s Chief AI Scientist Yann LeCun said OpenAI was “drunk on its own hype,” while Google DeepMind CEO Demis Hassabis called the episode “embarrassing.” OpenAI researcher Sébastien Bubeck later conceded that the model had “only rediscovered existing results,” effectively ending the incident as a self-inflicted debacle.

Further turmoil surrounded GPT-5’s launch presentation in August, when comparison charts were discovered to have exaggerated performance metrics. The distorted visuals triggered accusations of “chart manipulation,” prompting CEO Sam Altman to apologize, attributing the issue to “a late-night mistake by a staff member.” However, critics argued it reflected a deeper intent to inflate results.

Amid these missteps, OpenAI has now made a controversial “forbidden” move. CEO Sam Altman announced that beginning in December, the company will allow verified adult users to generate erotic content. “Adults should be treated as adults,” he said, framing the decision as a balance between “freedom of expression and mental well-being.”

Industry observers see the move as another attempt at user lock-in—an effort to revive engagement by accommodating demand that had migrated elsewhere under prior content restrictions. As strict safety filters and censorship drove user dissatisfaction, OpenAI now appears intent on absorbing that traffic back into its official ecosystem. Across online communities, prompt-sharing techniques for bypassing filters have already proliferated, while some users have shifted to alternative chatbots such as Character.AI or Grok.