Google’s AI reorganization makes for a dramatic headline. Demis Hassabis moved away from day-to-day leadership of DeepMind, Koray Kavukcuoglu took operational control, and several senior AI figures left.
But for a company deciding what to build on, the more useful fact is less glamorous. Google’s flagship Gemini 3.5 Pro missed its planned June launch, remained in partner testing in late July, and Reuters reported in August that internal performance concerns had pushed the release roughly two months behind plan.
Unreleased AI capability should not become a dependency in a live business plan.
UAE CIOs, founders, developers, procurement teams, and companies standardizing workflows around Gemini.
Use Google's available models where they work today while keeping high-risk workflows portable across providers.
Reported internal delays and leadership tensions do not establish how the unreleased model will ultimately perform.
Google says Gemini 3.5 Pro is in testing and Gemini 4 is already in training, with a stronger push toward commercialization.
List every workflow that depends on a future Gemini feature and create a working fallback using capability available today.
Organizations that separate delivered capability from roadmap promises and keep switching costs manageable.
Any business already evaluating or using Google AI products can apply the procurement and fallback checks in this guide.
Public release timing, API pricing, enterprise terms, migration requirements, and measurable coding performance.
That is not proof that Gemini is failing. Google is still shipping models, serving huge enterprise demand, and training Gemini 4. It is a reminder that AI roadmaps move quickly enough that businesses should buy what exists, not what a keynote implies will exist next quarter.
What Actually Changed at DeepMind
Alphabet announced the leadership overhaul on August 5. Hassabis became DeepMind chairman and Alphabet chief scientist, while Kavukcuoglu took operational leadership. Reuters later reported that the reorganization consolidated more authority around the executive already overseeing Gemini development and commercialization.
Several senior AI figures also left. That matters for the lab, but org charts are poor buying guides. A customer cannot put a management structure into production. It can only deploy an API, model, product, and set of terms that exist.
That distinction is easy to lose when the AI race is covered like a sports table. The geopolitical layer we examined in The US Wants AI Partners to Pick a Side makes vendor dependency even more important.
The Delay Is More Useful Than the Drama
Google originally indicated that Gemini 3.5 Pro would arrive in June. On July 21, Reuters reported that the flagship remained delayed while Google released lighter Gemini models instead. Google said the Pro model was still being tested with partners and would come soon.
On its July earnings call, Google confirmed that Gemini 3.5 Pro was still in testing and said Gemini 4 had begun its most ambitious pre-training run yet.
Reuters’ August reporting added that internal tests had exposed weaknesses, including coding performance, and described a roughly two-month delay. Those details are attributed reporting, not a public Google benchmark disclosure.
What This Means for UAE Buyers
A UAE company choosing an AI platform does not need to decide which lab will win. It needs to know whether the specific capability required for a workflow is available, stable, priced, supported, and permitted under the organization’s data rules.
That is especially important when AI is entering customer service, legal review, finance, coding, and government-facing workflows. A promised model can be late. A preview can change. An API can be repriced.
Our guide, The Best AI Model May Be the Wrong Business Choice, argues for routing work to the model that fits each task. The same principle protects a company from becoming dependent on a future release.
Buy the Product, Not the Roadmap
For each AI workflow, write down the minimum capability you require today. Then identify the model and version currently meeting it.
If a vendor promises a coming feature that would materially improve the workflow, treat it as upside. Do not make the business case depend on it until the feature ships under terms you can accept.
That sounds conservative. It is also how normal enterprise software procurement works. AI’s release speed made many teams forget the rule.
Google Still Has a Very Large Hand
The delay should not be read as evidence that Google’s AI business has stalled. Google said in July that more than nine million developers were building monthly with its models and that its model APIs were processing about 22 billion tokens per minute. Those are company-reported figures, but they illustrate the scale of the existing platform.
Google is also shipping lighter models, operating its own TPU infrastructure, and embedding Gemini across Search, Workspace, Cloud, and developer products. The company has several ways to compete even when a single flagship release slips.
That is why a binary ‘ahead’ or ‘behind’ frame is not useful to a UAE buyer.
The Procurement Checklist
Ask five questions before tying a workflow to any frontier model. Is the capability generally available? Is the exact model version named in your architecture? What happens when that version is retired? Can you move the workload without rebuilding everything? Which data and logs leave your environment?
Then add one more: if the next flagship is delayed by three months, does your business plan still work?
The answer should be yes. The valuation assumptions around frontier labs in Anthropic’s IPO math depends on future revenue are another reminder that the entire sector is moving on forecasts as well as delivered products.
Sources
• Reuters: August reporting on the DeepMind leadership reshuffle and reported Gemini delay – https://www.reuters.com/world/inside-google-executive-moves-that-led-its-big-ai-reshuffle-2026-08-12/
• Reuters: July reporting on lightweight Gemini releases and the delayed flagship – https://www.reuters.com/business/google-updates-lightweight-gemini-models-flagship-still-delayed-2026-07-21/
• Google: Alphabet Q2 2026 earnings remarks confirming Gemini 3.5 Pro testing and Gemini 4 training – https://blog.google/company-news/inside-google/message-ceo/alphabet-earnings-q2-2026/
Robius.news — Dubai, UAE — 2026 | Built to be first. Built to be trusted.
