OpenAI has launched the o3-pro artificial intelligence model, which emphasizes increased reliability and enhanced tool integration for enterprise applications, albeit at the cost of slower response times. The new model is designed to deliver more accurate and detailed answers, making it particularly appealing to developers and businesses that prioritize precision. The o3-pro model, which replaces the previous o1-pro, has been reported to take significantly longer to generate responses; for instance, one user noted a response time of three minutes for a simple greeting. OpenAI stated that the o3-pro is particularly suited for complex inquiries where accuracy outweighs the need for speed. Pricing for the o3-pro model is notably higher than its predecessor, costing $20 per input and $80 per output, compared to the o3 model’s new reduced rates of $2 and $8, respectively. The launch of o3-pro comes as OpenAI reports reaching three million business users, with a notable 50% increase in enterprise clients since February, as well as hitting $10 billion in annual recurring revenue, just under three years after the launch of its ChatGPT chatbot. This revenue includes sales from consumer products, business services, and application programming interfaces, while excluding licensing revenue from Microsoft and one-time deals.
In the previous year, OpenAI recorded approximately $5.5 billion in annual recurring revenue. Despite this rapid growth, the San Francisco-based startup reported a loss of around $5 billion last year. OpenAI aims for a remarkable $125 billion in revenue by the year 2029, according to confidential sources. As of late March, OpenAI supports 500 million weekly active users and has three million paying business users, a significant increase from two million reported in February.
To meet this demand, OpenAI has entered a significant partnership with Alphabet’s Google Cloud service to enhance its computing capabilities, marking a notable shift in the competitive landscape of artificial intelligence. This collaboration, finalized in May 2025, allows OpenAI to diversify its computing sources, reducing its reliance on Microsoft, which has been its primary infrastructure provider. Analysts at Scotiabank described the arrangement as somewhat surprising, underscoring the willingness of both companies to overlook their competitive rivalry to meet massive computing needs. Google’s cloud sales accounted for 12% of Alphabet’s total revenue in 2024, indicating the importance of cloud services in the company’s overall strategy as it seeks to balance its enterprise and consumer business segments.
Why do we care?
OpenAI’s o3-pro is not just a model—it’s a signal. The company is clearly betting that enterprises will pay top dollar for precision, and that they’ll accept trade-offs in speed and cost if the output is measurably better. This creates a clear service design opportunity for IT services providers: build AI offerings around use cases where latency is tolerable but stakes are high—such as contract review, compliance automation, or executive decision support.
At $20 per input and $80 per output, o3-pro’s costs are prohibitively high for most SMBs and many midmarket clients. This locks out smaller players and could deepen the divide between tech “haves” and “have-nots.” Adding in the legal risks around data sourcing and the complexity of multi-cloud orchestration, the bar for AI entry is rising quickly—especially for smaller IT providers. That’s a strategic opportunity: MSPs who can simplify this complexity and provide AI that’s compliant, cost-aligning, and contextualized will quickly win client trust. Additionally, “More accurate and detailed answers” is vague. Without published benchmarks, testing transparency, or real-world validation, it’s unclear how o3-pro performs compared to Claude or Gemini in enterprise contexts.
At the same time, diversification of OpenAI’s cloud partnerships should prompt MSPs to review their own stack decisions. Just as OpenAI is hedging its infrastructure bets, service providers should reassess their AI vendor dependencies and look for multi-cloud, model-agnostic architectures. The game is no longer about which AI model is best overall—it’s about which model is best for each use case, and how you deliver that to clients.
OpenAI may be winning on revenue, but reliability, profitability, latency, and access equity remain unresolved. That’s where IT providers can step in—to optimize, contextualize, and operationalize AI.