On the same day, two giants each made their move.

Author and source: 0x9999in1, ME News

TL;DR

• On May 14, 2026, OpenAI CEO Sam Altman announced that enterprise users switching to Codex within 30 days will receive two months of free access, along with a one-click migration tool.
• On the same day, Anthropic increased the weekly usage limit for Claude Code by 50% for Pro, Max, Team, and Enterprise users, valid until July 13.
• This is no coincidence—it’s a carefully planned head-to-head battle for the world’s most valuable user base: developers.
• The AI coding market has entered a "multi-strong competition" landscape: GitHub Copilot boasts 4.7 million paying users, Cursor has surpassed $2 billion in annualized revenue, and Claude Code leads in satisfaction (ranked as the most favored by 46% of respondents in JetBrains’ April 2026 survey).
• The real logic behind it: AI companies are burning cash not to achieve immediate profitability, but to lock in the largest possible user base before the IPO window opens.
• Subsidies will continue, but the era of "free lunch" is fading—true costs are far higher than the subscription price you see.

One day, two giants each make a move.

On May 14, something interesting happened in Silicon Valley.

Sam Altman posted directly on X, stating plainly: “Codex is the best AI coding product.” Then he shifted gears—enterprise customers switching over within a 30-day window will receive two months of full enterprise edition access.

It’s not just talk. OpenAI has also launched a migration tool that enables seamless transfer of settings, plugins, skills, infrastructure, projects, and conversation history.

It means: Don’t hesitate—pack your bags, and we’ve even arranged a moving company for you.

What about Anthropic? Almost right behind. Claude Code’s weekly quota is being increased by 50% for Pro, Max, Team, and seat-based Enterprise users, valid until July 13.

Why July 13? Two months. It almost exactly overlaps with OpenAI's free trial period.

This is not a coincidence.

II. Background: A Conflict Long in the Making

To understand why this "price war" has erupted now, let’s first look at a set of data.

Since early 2026, the AI Coding sector has become one of the fastest-growing verticals in the global tech industry. A January 2026 survey by JetBrains showed that 90% of developers use at least one AI programming tool regularly in their work. This is no longer a novelty—it’s infrastructure.

What about the market landscape?

• GitHub Copilot: 4.7 million paid subscribers, a 75% year-over-year increase, deployed in 90% of the Fortune 100 companies.
• Cursor: Annual revenue surpassed $2 billion in February 2026, doubling in three months; $50 billion valuation ahead of upcoming funding round.
• Claude Code: Leading significantly in satisfaction—46% of developers selected it as their "favorite tool" in JetBrains' April 2026 survey, compared to 19% for Cursor and just 9% for Copilot.
• OpenAI Codex: A late entrant, but rapidly rising by leveraging the ChatGPT ecosystem; launched a plugin system and usage-based pricing for enterprise solutions in April.

Here’s the question: Where does OpenAI’s Codex stand in this landscape?

The answer is—follower.

Claude Code won the reputation, Cursor won the revenue growth rate, and Copilot won the scale. What about Codex? Its advantage is the entire ecosystem behind OpenAI: ChatGPT, GPT-5, the financial firepower behind an $852 million valuation, and Sam Altman’s own “full commitment.”

That's why Altman said on this day: "We want to make trying it simple."

We know we’re not yet number one, but we’re going to make the cost of switching to us zero.

III. Anthropic's Counterattack: Not Defense, but "Building the Wall Ahead"

Anthropic's 50% increase appears on the surface to be a defensive move. But upon closer inspection, it reveals the subtle challenges Anthropic has faced over the past month.

Rewind the timeline:

• April: Anthropic began restricting third-party agent tools (such as OpenClaw and T3 Code) from calling Claude via subscription quotas, citing "prevention of abuse."
• May 6: Anthropic announced a computing agreement with SpaceX, securing all computing power from the Colossus 1 data center—exceeding 300 megawatts and 220,000 NVIDIA GPUs (including H100, H200, and the latest GB200).
• May 13: Claude Code quota increased by 50%
• May 14: Announcement of the new Agent SDK billing policy—starting June 15, third-party programmatic calls will use a separate credit pool, fully isolated from regular chat quotas.

Can you see the rhythm?

Anthropic is simultaneously bolstering its computational resources (through its partnership with SpaceX) and redefining the rules around “who gets how much.” Third-party tool usage is now isolated into a separate credit pool, effectively reclaiming previously exploited inference resources and reallocating the saved quotas as a 50% increase for direct paying users.

With this cut, some estimates suggest that subsidy inference volumes from third-party tools were reduced by 25 to 40 times.

So when OpenAI said, "Come to me, two months free," Anthropic responded: "Don't leave. I'll give you more."

At the same time, this implies that my cost structure is being adjusted, so the price you pay in the future will be clearer—and higher.

IV. The Essence of Price Wars: It’s Not About Lowering Prices, It’s About Seizing "Lock-in Effects"

Let's take a break.

Is OpenAI's "two months free" really free? Of course not.

According to Verdent.ai’s estimates, the actual cost of using Codex Enterprise is approximately $100 to $200 per developer per month. With two months free for a 50-person team, this amounts to OpenAI providing a one-time "subsidy" of $10,000 to $20,000.

The amount itself is not large. But what it buys is expensive—the inertia after the migration is complete.

Your settings have been migrated. The plugins have been adapted. The project is running. Your chat history remains intact. Move everything back in two months? The migration cost far exceeds the savings from "giving you two more months."

This is the classic "subsidize first, monetize later" model. Ride-hailing apps did it, food delivery services did it, and now it’s AI programming tools’ turn.

Anthropic’s strategy is slightly different. Instead of “free entry,” it offers “more value for the same price”—maximizing the experience you already pay for, whether it’s $20/month for Pro or $200/month for Max. Since you’re already here, I’ll make your experience even better, so you won’t be tempted away by “two months free” offers from competitors.

Two strategies: one offensive, one defensive. But the goal is the same: lock users into their own ecosystem before the IPO window opens in the second half of the year.

Five: IPO Countdown: Where Is the Funding Coming From?

OpenAI: Completed the largest private funding round in human history in April 2026—$122 billion—resulting in a post-money valuation of $852 billion. Investors included SoftBank ($30 billion), NVIDIA ($30 billion), and Amazon ($50 billion). Annualized revenue surpassed $25 billion during the same period. IPO target: Second half of 2026.

Anthropic: Completed its Series G funding in February 2026—$30 billion, post-money valuation of $380 billion. Annualized revenue surged from $9 billion at the end of 2025 to over $30 billion by April 2026. IPO rumors also point to 2026.

Together, the two companies have over $150 billion in fresh capital.

Compared to this level of capital, “two months free” and “50% higher limits” are negligible. But their signaling value far outweighs their financial impact—they tell Wall Street analysts: our user base is growing, our ecosystem is expanding, and we deserve that IPO valuation.

That's why neither company cares about short-term costs.

For them, each enterprise client that migrates is just a footnote written to an investment bank.

Six: The Real Dilemma for Developers: What Was Cheap Is Getting More Expensive

Is a price war good for developers?

In the short term, certainly. Why not take advantage of free trials and increased limits?

However, in the long term, the signal is not particularly optimistic.

A developer on Medium shared their data: their personal API costs have reached $500 to $2,000 per month. A engineering lead on Substack stated verbatim: “We’re spending $1,400 per month on three different AI programming tools. My boss said: pick one. By Friday.”

The issue is that the "true cost" of AI programming tools is coming to light.

According to analysis by noqta.tn, the actual cost of providing the service behind the $20/month subscription price may be two to five times higher—that is, the company is losing $40 to $80 per paying subscriber each month.

This subsidy cannot last forever. Anthropic’s restrictions on third-party tool access and OpenAI’s shift in April from a fixed-seat model to pay-per-token pricing are both precursors to cost recovery.

PYMNTS put it well: “For the past two years, you could get unlimited AI programming assistance for $10 a month. That deal is over.”

The irony of this price war is that both sides aren’t competing over who’s cheaper, but rather who can attract more users before raising prices.

Seven: The Darkest Hour for Third-Party Players

In this showdown between the two giants, the biggest loser is neither OpenAI nor Anthropic.

It is Cursor. It is any intermediary product that relies on their model capabilities but does not have its own infrastructure.

Think about it: if OpenAI offered enterprises two months free plus a one-click migration tool; if Anthropic maximized Claude Code’s credits—why would a business still pay an extra layer of fees to Cursor?

Cursor certainly has its own barriers: an IDE experience, real-time inline editing, and a network effect from over 360,000 paying users. It is projected to achieve an annualized revenue of over $6 billion by the end of 2026. This is a healthy company.

But Codex’s latest moves—plugin system, desktop app, browser integration, PR review, and 90+ plugins—are essentially layering on the very differentiators that Cursor has established.

Claude Code as well. Native terminal-based deep reasoning and a 200K standard context window (Opus 4.6’s 1M token beta)—it’s eroding Cursor’s market space from another angle.

Attacked from both sides.

The story of Cursor's $5 billion valuation rests on the premise that it is a "model-agnostic" middleware layer. But when model providers themselves enter the market with end-to-end products, the value of such a middleware layer must be reevaluated.

Eight: Judgment: The Three Endgames of This War

First, in the short term, developers are the winners. Free trials will become more common and allowances will grow larger. Before the second half of 2026, the benefits you can gain may exceed what you’ve received over the past two years combined.

Second, long-term pricing will rise significantly. Once the IPO is completed and valuation is locked in, subsidies will inevitably fade. At that point, the true monthly cost of AI coding tools will stabilize at $100–$200 per user, rather than the current illusion of $20. Token-based billing will become the norm.

Third, the market won’t be reduced to just two players. GitHub Copilot has Microsoft’s distribution channels. Cursor offers an unparalleled editor experience. Open-source solutions (OpenCode, Crush) are rising in the $5–$45 price range. AI coding tools won’t follow a “winner-takes-all” model like search engines; instead, they’ll resemble the IDE market—multiple players coexisting long-term, each with their own stronghold.

But one thing is certain: after the two free months are over, the bill will arrive.

And it will be more expensive than you think.

Nine, Epilogue

When Sam Altman posted on X, his wording was carefully crafted.

He didn’t say “Codex is the cheapest.” He said “Codex is the best.” Then he added: “We want to make trying it simple.”

Only someone unsure of their product’s strength would say that. True market leaders don’t need to offer you two months of free trial—you’ll come on your own.

But this is precisely what makes OpenAI so formidable—it isn’t afraid to admit it’s not yet number one, but it has the resources, patience, and determination to invest until it becomes number one.

A company valued at $852 billion is giving you two months free. Do you think it's losing money?

No. It is investing.

The investment target is you—and you, in its IPO pitch deck, are just a number.

Source citation

1. OpenAI, "Codex now offers pay-as-you-go pricing for teams," openai.com, May 2026.
2. OpenAI, "Introducing the Codex app," openai.com, May 2026.
3. Anthropic, "Higher usage limits for Claude and a compute deal with SpaceX," anthropic.com, May 6, 2026.
4. Reuters, "OpenAI tops $25 billion in annualized revenue," reuters.com, March 5, 2026.
5. Bloomberg via Yahoo Finance, "Anthropic Tops $30 Billion Run Rate, Seals Broadcom Deal," finance.yahoo.com, April 2026.
6. OpenAI, "Scaling AI for Everyone" (funding announcement: $122B at $852B valuation), openai.com, April 2026.
7. Anthropic, "Anthropic raises $30 billion in Series G funding at a $380 billion post-money valuation," anthropic.com, February 12, 2026.
8. IdeaPlan, "AI Coding Assistant Market Share 2026" (citing JetBrains April 2026 survey data), ideaplan.io, 2026.