Recommended Free Tools
The headline refers to a June 27, 2023 VentureBeat interview with Cohere co-founder and CEO Aidan Gomez and president Martin Kon. It followed Cohere’s announcement of a $270 million financing round involving Nvidia, Oracle, Salesforce Ventures and other investors. VentureBeat reported that the round valued Cohere at more than $2 billion, often described at the time as approximately $2.1 billion. The interview’s larger argument was that an independent, enterprise-focused model provider could serve customers across cloud environments while addressing privacy, customization and deployment concerns.
The financing figures and executive comments are historical facts and attributed positions from 2023—not evidence of Cohere’s current valuation, product lineup, pricing, regulatory status or technical performance.
What the 2023 funding announcement actually meant
VentureBeat reported that Cohere’s $270 million round included Nvidia, Oracle and Salesforce Ventures. The company had been founded in 2019 by Aidan Gomez, Ivan Zhang and Nick Frosst. At the time of the interview, the financing placed Cohere’s reported valuation above $2 billion.
The investor list mattered because it combined capital with relationships across the AI infrastructure and enterprise technology markets. Nvidia represented access to the accelerator ecosystem; Oracle represented enterprise cloud infrastructure and security relationships; Salesforce Ventures added another enterprise-software connection. Those strategic benefits were part of the appeal described by Cohere’s leadership.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →#1 Best Overall
Participation in a financing round does not mean that Nvidia or Oracle acquired Cohere, controlled it, guaranteed exclusive distribution or committed every customer to a particular cloud. An equity investment, commercial partnership, cloud-distribution agreement and hardware-supplier relationship are different arrangements. The interview presented Nvidia and Oracle as strategic and financial supporters of an independent provider.
Read the original interview at VentureBeat.
Why Nvidia and Oracle were strategically important
Nvidia: compute and an ecosystem, not simply a venture check
Nvidia supplies the accelerators and systems used to train and run many modern AI models. Gomez linked Nvidia’s importance to the broader compute ecosystem and noted that Nvidia technology was available through multiple cloud providers. That distinction supported Cohere’s argument that using Nvidia hardware did not require becoming tied to one hyperscaler.
For an enterprise model company, an Nvidia relationship can potentially help with optimized inference, access to specialized infrastructure and credibility with customers planning GPU deployments. None of those possibilities establishes a guaranteed allocation, preferential pricing or exclusive access for Cohere customers.
Oracle: enterprise infrastructure and data controls
Kon connected Oracle to enterprise infrastructure, security and data-protection priorities. Existing Oracle customers may already have procurement processes, identity systems, databases, network controls and compliance reviews built around Oracle Cloud Infrastructure.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallThat relevance is different from saying Cohere was an Oracle-only service. The executives instead described a provider that could work across cloud environments, including Oracle’s.
What “cloud-agnostic” meant in Cohere’s pitch
Cloud-agnostic means software can be deployed across multiple cloud environments. It does not mean cloud-independent: a model still needs cloud capacity, chips, networking, storage, monitoring and other infrastructure.
Gomez and Kon said Cohere wanted its models to move among clouds and, in some situations, operate across them simultaneously. For enterprise buyers, that approach can address several practical concerns:
- Negotiating leverage: customers may avoid making one hyperscaler their only route to an important model service.
- Data residency and protection: workloads can be placed where regional, contractual or internal data controls require.
- Private deployment: organizations may seek private-cloud, virtual-private-cloud or otherwise controlled environments.
- Existing infrastructure: a company can use its current cloud relationships instead of rebuilding every system around a new provider.
Portability does not erase switching costs. Integrations, retrieval pipelines, prompt templates, monitoring, security reviews, contracts, application behavior and staff expertise can all create lock-in. A buyer should test migration rather than treat a “multicloud” label as proof of effortless portability.
Gomez contrasted Cohere’s positioning with OpenAI’s enterprise offering by citing dependence on Azure. That was his 2023 comparison, not a complete or current assessment of every provider’s deployment option.
Gomez’s response to Geoffrey Hinton’s AI-risk warnings
Gomez and co-founder Nick Frosst had connections to Google Brain, and the interview described Geoffrey Hinton as both a respected researcher and a Cohere investor. Hinton had recently left Google and spoken publicly about serious AI risks.
Gomez said he took Hinton’s expertise seriously but placed greater emphasis on harms that were already visible or could arise soon. He identified:
- synthetic media and false information;
- bias and hallucinations;
- deploying unreliable systems in high-stakes settings;
- poor governance of models already in use; and
- the social consequences of inappropriate deployment.
His emphasis differed from Hinton’s focus, as characterized in the interview, on longer-term or existential risks to humanity. Gomez did not present this as proof that Hinton was wrong. His position was that risk management should cover both near-term operational harms and longer-horizon catastrophic possibilities.
Synthetic data, model collapse and the future of LLMs
The model-collapse concern
Researchers had warned that repeatedly training models on generated material can degrade performance. If synthetic outputs replace diverse, high-quality human or real-world data, later generations may lose information, diversity or accuracy and amplify artifacts already present in the generated material.
Gomez’s qualification
Gomez treated model collapse as a problem associated with particular ways of collecting and recycling synthetic data, not as an unavoidable property of every use of generated data. Proper filtering, provenance tracking, diversity checks and validation against external reality are material safeguards.
His longer-term prediction
Gomez also predicted that carefully used synthetic data could help models discover useful knowledge, improve reasoning or move beyond the limits of publicly available human-generated text. That is a thesis from 2023, not an established consensus or verified forecast. Synthetic data can expand training examples, but indiscriminate recursive training can also magnify errors and bias.
Rank #4
The enterprise operating problem: releases, testing and drift
Gomez said customers needed education about where language-model applications were and were not appropriate. In the interview, Cohere recommended customer-specific test sets and continuous benchmarking rather than automatic adoption of every new model release.
That advice matters because Gomez said Cohere was releasing models approximately weekly at the time. A frequent release cadence can improve capabilities while changing outputs, latency, safety behavior or tool interactions. Enterprises should make model updates a controlled change-management process:
- Define acceptance tests: include representative prompts, prohibited requests, required citations, latency limits and business-specific quality thresholds.
- Pin the production version: do not allow an unreviewed provider update to change a critical workflow automatically.
- Evaluate a candidate release: run the same test set against the current and proposed versions and inspect regressions, not just average scores.
- Monitor after deployment: track hallucinations, refusals, bias complaints, drift in user inputs and changes in cost or latency.
- Keep a rollback path: preserve the previous version, prompts, retrieval configuration and data-processing settings.
These controls also force questions about data provenance, retention, permission to use training or customer data and the rights attached to material supplied to the system.
How Cohere framed enterprise models versus open models
Gomez said open source was advancing quickly but argued that managed enterprise providers offered a different value proposition: frequent updates, a close customer-feedback loop, influence over model direction, support and deployment controls. That was Cohere’s positioning argument, not an independent benchmark showing superiority.
| Question | Managed enterprise model | Open or self-hosted model |
|---|---|---|
| Deployment effort | Usually lower; provider or partner operates much of the service | Higher; the customer operates more of the stack |
| Control over weights | Usually limited | Generally greater when weights are available under usable terms |
| Update cadence | Provider-controlled unless version pinning is offered | Customer-controlled, with more maintenance responsibility |
| Infrastructure burden | Provider-managed or shared | Customer-managed hardware, hosting and optimization |
| Portability | Depends on interfaces, contracts and deployment options | Can be greater, but tooling and license restrictions vary |
| Data governance | Requires review of provider retention, training-use and regional terms | Requires securing the entire self-managed stack |
| Cost profile | Usage or contract charges plus integration and monitoring | Infrastructure, engineering, licensing and operations costs |
Open models can offer inspectability, deployment control and data-sovereignty advantages. A managed service can reduce engineering work and provide support. Licenses differ substantially, and “open” does not automatically mean unrestricted commercial use or easy operation.
Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Best Value
What an enterprise buyer should examine
If evaluating the strategy Cohere described in 2023, separate marketing claims from contractually documented capabilities. Ask:
- Can the model run in the clouds and regions the organization actually uses?
- Are residency, retention, access and provider-training terms explicit?
- Can customers customize terminology, retrieval and workflows without surrendering control of sensitive data?
- Are version pinning, rollback and release notices available?
- What customer-specific evaluations, support commitments and service levels are documented?
- What are the exit costs if the provider, model or commercial terms change?
- Does total cost include inference, integration, data preparation, monitoring and staff time—not just token charges?
- Does strategic participation by infrastructure companies create useful access, or perceived conflicts that procurement and risk teams must assess?
Portability, privacy and customization may be valuable, but they should be verified in architecture reviews, security questionnaires, pilot tests and the final agreement.
Transparency and the limits of the 2023 record
Gomez said Cohere tried to answer customer questions about training data while protecting intellectual property. He discussed screening toxic material, data provenance, permission to train and compliance with robots.txt as characterized in the interview.
Those statements should be read as company assertions from 2023. They do not establish that every Cohere model or dataset is fully transparent, copyright-safe or legally settled. A buyer still needs current documentation covering training-data disclosures, licenses, customer-data use, retention and applicable law.
Free tools Windows power users keep installed
One-click scans. No signup required.
Regulation: why the date matters
The interview discussed the then-draft EU AI Act. That was historical context in June 2023. It should not be used as evidence of Cohere’s current compliance, the present status of European law or today’s obligations for general-purpose AI providers. Current legal conclusions require current legal sources and a review of the specific deployment, jurisdiction and provider terms.
What remains useful—and what remains unproven
The interview’s durable business insight is that enterprise buyers care about where models run, how data is governed, how systems are evaluated and whether a provider can be replaced. Those questions remain relevant regardless of which model is technically strongest.
Its less certain claims are the forecasts: that synthetic data will let models move beyond human-generated knowledge, that particular approaches can avoid collapse, and that a cloud-agnostic provider will materially reduce lock-in for every customer. Those propositions require evidence beyond the 2023 interview.
For personal-finance and business readers, the practical lesson is straightforward: treat the $270 million round and reported valuation as historical financing news, treat the cloud and safety arguments as executive positions, and evaluate any current purchase using present contracts, technical tests, operating costs and legal requirements.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsQuick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
