Security worries are putting on-premises tech back in the discussion
Strategic adviser Itay Sagie says AI fraud, private AI workloads and quantum risk are pushing some companies to reconsider cloud-only setups.
By Theo Nakamura · Staff Writer
· 3 min read
Some enterprise buyers are asking again for on-premises technology, according to strategic adviser Itay Sagie, a guest contributor to Crunchbase News. For retail investors watching cloud, cybersecurity and enterprise software names, the signal is less about a full retreat from cloud and more about how security concerns may shift tech spending.
On-premises systems are hardware and software that a company runs in its own facilities or controlled environment, rather than relying fully on third-party cloud providers. Sagie said a PBX vendor, meaning a provider of private business phone systems, recently told him customers are asking for those systems again.
Sagie’s view is that cloud migration still has a strong case. Over the past decade, companies used cloud services to move faster, scale without buying as much hardware up front and modernize without rebuilding infrastructure themselves. Startups could launch products without owning servers, while large enterprises could upgrade systems through outside providers.
The change, Sagie argued, is that advances in artificial intelligence and future computing risks are making some decision-makers more uneasy about where sensitive systems and data sit.
AI fraud is widening the risk surface
Sagie wrote that cloud providers are often more secure than systems a company could build on its own. The issue, in his view, is that AI is making fraud harder to spot and easier to produce at speed.
He pointed to more convincing phishing messages, fake invoices and voice impersonation. A request that appears to come from a chief financial officer or chief executive, including by voice, is becoming more believable as AI voice cloning improves, Sagie wrote.
That shifts the security conversation beyond servers. Sagie said companies now have to think about identity systems, software-as-a-service tools, application programming interfaces, employee processes, access permissions, contractors and support portals. An application programming interface, or API, is the connection that lets software systems exchange data or instructions.
For communications, payments, identity and customer data, Sagie said some companies may place a higher value on direct control. He cautioned that on-premises systems do not automatically make a company secure, but said they can reduce reliance on outside platforms and give businesses clearer control over high-risk systems.
Private AI could change the economics
Sagie also said enterprise AI is moving from testing into production. During pilots, cloud AI tools can be attractive because companies can experiment without buying graphics processing units, managing models or hiring large infrastructure teams.
Production use can look different. Sagie said the most valuable business AI often depends on proprietary data, including contracts, source code, customer records, financial reports, support tickets, security logs, medical files and internal messages.
For those workloads, Sagie said on-premises or private AI infrastructure may become more attractive because models can operate closer to the data and access can be controlled more tightly. He also said retention, compliance and audit requirements may be easier to manage in private environments.
Cost is another factor. Sagie said token pricing, where customers pay based on AI usage, is convenient for pilots but can become expensive when thousands of employees or customers use AI daily. For stable, high-volume workloads, he argued that owning or controlling infrastructure may cost less than paying for each interaction over time.
Quantum risk is entering planning
Sagie said quantum computers are not breaking enterprise encryption today, but the risk is already part of serious security planning. Encryption is the process of scrambling data so it cannot be read without the right key.
His concern is long-term data exposure if future commercial quantum computers can defeat current encryption methods. Sagie said that issue matters most for organizations holding sensitive data for long periods, including banks, healthcare providers, telecom companies, governments, defense-related organizations and infrastructure providers.
Sagie’s conclusion is cautious: on-premises technology may not be the best answer for every risk, but some buyers may increasingly perceive it as safer. That perception alone, he said, could support more demand for older on-premises strategies.
This story draws on original reporting from Crunchbase News.