OpenAI API Price Hikes, Meta Layoffs, SAP Acquisition: The AI Week Unpacked

Google I/O ran Tuesday and Wednesday and the headlines were about Gemini models and smart glasses. The more important story landed Monday morning, before the keynote started. Blackstone committed $5 billion to a joint venture with Google to bring TPU compute-as-a-service to market, the first serious structural challenge to Nvidia’s grip on AI infrastructure in years. The same week, OpenAI was projecting a $14 billion loss on $25 billion in annualized revenue after doubling its API prices. Meta began cutting 8,000 employees on Wednesday to free up budget for AI capital expenditure. SAP agreed to pay over $1.18 billion to acquire a small German AI lab most people had never heard of. The week was not about model releases. It was about who controls the infrastructure layer, what that control costs, and which smaller players get squeezed or lifted by the answer.

Blackstone just handed Google $5 billion to break Nvidia’s hold on AI compute

On May 18, Blackstone announced a $5 billion equity commitment to a joint venture with Google that will offer data center capacity, networking, and Google’s custom Tensor Processing Units as a compute-as-a-service product. The venture, internally called N1, expects to bring its first 500 megawatts of capacity online in 2027 with plans to scale significantly from there. Google also used I/O to announce its eighth-generation TPU architecture, TPU 8t for large-scale AI training and TPU 8i for inference workloads, alongside Gemini 3.5 Flash and the new Gemini Omni multimodal model family.

The infrastructure story is the one that matters for smaller players. Tensor Processing Units (TPUs) are custom chips designed specifically for AI workloads, optimized for training and inference tasks rather than the general-purpose computing that GPUs were originally built for. Nvidia has dominated AI compute because it got there first and its CUDA software ecosystem created switching costs that kept customers locked in. The Blackstone-Google venture is the most credible attempt yet to offer an at-scale alternative. It will not displace Nvidia overnight. But it introduces price competition into a market that has had very little of it.

For smaller AI companies and startups, the current compute environment is brutal. Nvidia GPU access is expensive, waitlisted, and controlled by the hyperscalers who resell it at a premium. A new compute-as-a-service option with 500 megawatts of capacity coming online in 2027 is a meaningful development for any company that builds on rented infrastructure. The cost of running AI workloads is a primary constraint on smaller AI builders right now. Anything that applies downward pressure on that cost improves the economics of the entire ecosystem below the hyperscaler level.

OpenAI doubled its API prices

The five largest technology companies plan to spend between $630 billion and $700 billion on AI infrastructure in 2026. Amazon leads at $200 billion, Alphabet at $175 to $185 billion, Meta at $125 to $145 billion, Microsoft tracking toward $120 billion, and ByteDance discussing capex of up to $70 billion underwritten by its $50 billion in 2025 profits. Nvidia is spending $100 to $150 billion annually on its Taiwan supply chain alone, up from $10 to $15 billion four years ago.

The cost of that infrastructure build is flowing downstream. OpenAI launched GPT-5.5 at $5 per million input tokens and $30 per million output tokens, double the pricing of GPT-5.4, which launched just six weeks earlier. The company also deprecated its fine-tuning API and GitHub Copilot is shifting to usage-based token billing effective June 1. OpenAI is projecting a $14 billion loss in 2026 despite 900 million weekly active users and $25 billion in annualized revenue. The pricing increases are not arbitrary. They reflect real cost pressure from a capital expenditure cycle that is running ahead of revenue growth.

For smaller companies building products on top of frontier AI APIs, this is a structural cost problem. Every repricing event by OpenAI, Google, or Anthropic hits their margins directly. The companies best positioned in this environment are the ones that have either diversified across multiple model providers, built on open-source models where the compute cost is more controllable, or found a niche where the value they deliver can absorb higher input costs. The ones most exposed are the thin-wrapper businesses that built on a single API at a price point that no longer exists.

Meta cut 8,000 jobs to pay for AI

Meta began cutting approximately 8,000 employees on May 20, representing 10% of its total workforce, while simultaneously closing 6,000 open roles and redeploying 7,000 employees into AI-focused positions. LinkedIn cut more than 600 jobs the same week, with LinkedIn CEO Daniel Shapero citing the need to reinvent how the company works and redirect investment toward infrastructure and long-term priorities. Meta raised its 2026 capital expenditure guidance to between $125 billion and $145 billion in the same reporting period. JPMorgan downgraded Meta shares, flagging a more challenging path to returns relative to competitors.

The pattern is consistent across Big Tech. Headcount reduction is the funding mechanism for AI capital expenditure. Companies are making an explicit bet that AI-augmented employees generate more output than a larger, less AI-enabled workforce. That bet may or may not pay off for the companies making it. For the software ecosystem around them, it creates a very specific demand signal.

Enterprises replacing headcount with AI tools need software to do it. Workflow automation, AI productivity platforms, HR technology, and enterprise process management tools are all positioned to benefit from a restructuring wave of this scale. The opportunity does not sit with the hyperscalers doing the cutting. It sits with the smaller software companies that can land inside the workflows being redesigned. Niche enterprise AI software operators with a clear productivity ROI story are the most fundable category in the market right now, and the Meta and LinkedIn announcements just made that case louder.

SAP paid $1.18 billion for a startup nobody had heard of

SAP agreed to acquire Prior Labs, a Freiburg-based startup developing tabular foundation models, and committed over $1.18 billion over four years to transform it into a globally leading frontier AI lab specializing in structured business data prediction. The deal is subject to regulatory approval. Prior Labs is not a household name. That is precisely the point.

SAP is the world’s largest enterprise resource planning software vendor, with deep penetration across manufacturing, logistics, finance, and supply chain operations at major corporations globally. What it does not have is a frontier AI capability built for structured tabular data, the kind of data that sits inside ERP systems. Tabular foundation models (TFMs) are AI models trained specifically on structured rows-and-columns data rather than text, designed to make predictions and surface patterns in the kind of operational data that enterprise software runs on. That is a narrow, specific capability. SAP paid over a billion dollars for it.

The acquisition signals where enterprise software incumbents are looking. They are not trying to build general-purpose large language models. They are acquiring specialized AI capabilities that plug into existing enterprise data infrastructure and deliver measurable operational outcomes. For smaller AI companies with narrow, defensible capabilities in enterprise data, security, compliance, or process automation, the SAP and Prior Labs transaction is a direct read on acquisition appetite. The buyers are not looking for another ChatGPT. They are looking for something that makes their existing product meaningfully better for the customers they already have.

What to Watch

GitHub Copilot token billing goes live June 1: the shift from request-based to usage-based pricing changes the cost structure for every developer team using Copilot. Watch for enterprise adoption data in the weeks following the transition and whether smaller development teams begin migrating to cheaper alternatives.

OpenAI IPO signals: OpenAI reported $2.6 billion in monthly revenue and 900 million weekly active users this week. An IPO later in 2026 has been flagged as a possibility. Any formal filing or timeline announcement would be the most significant tech capital markets event of the year and would reset valuations across the AI software sector.

Nvidia earnings: with the Blackstone-Google TPU venture announced and Big Tech capex at record levels, Nvidia’s next earnings report is the clearest read on whether infrastructure demand is being met or whether the supply constraint is still driving pricing power. Any softness in data center revenue growth would ripple through the entire AI infrastructure investment thesis.

EU AI Act enforcement: the first enforcement deadlines under the EU AI Act are approaching. Watch for any enforcement actions or compliance guidance that affects how smaller AI companies operating in European markets structure their products and disclosures.

Sources

Editorial Disclosure

This analysis is based entirely on publicly available information including press releases, earnings disclosures, and verified market data. Securities discussed include Alphabet Inc. (Nasdaq: GOOGL), Blackstone Inc. (NYSE: BX), Meta Platforms Inc. (Nasdaq: META), Microsoft Corporation (Nasdaq: MSFT), Amazon.com Inc. (Nasdaq: AMZN), NVIDIA Corporation (Nasdaq: NVDA), and SAP SE (NYSE: SAP). aktiego.com has not received any compensation from any company mentioned, their management, investor relations representatives, or any third party. No staff member or principal of aktiego.com holds a position in any security mentioned at the time of publication. All information is sourced from named public filings, press releases via named wire services, and verified data sources. OpenAI is a private company; revenue and loss figures cited are sourced from named published reports. Forward-looking commentary regarding IPO timelines, regulatory enforcement, earnings outcomes, and market developments is opinion only. Microcap and smaller-operator references are included for market context only and do not constitute investment recommendations. Coverage on aktiego.com is provided for informational and educational purposes only. aktiego.com is not a registered investment advisor. Nothing in this article constitutes financial, investment, or professional advice. Readers are encouraged to conduct their own due diligence and consult a qualified financial advisor before making any investment decisions. For more information please see our full DISCLAIMER.

OpenAI API Price Hikes, Meta Layoffs, SAP Acquisition: The AI Week Unpacked

Join our Mailing List