The Philippines needs to address gaps in power supply and connectivity alongside investments in data centers as businesses move artificial intelligence (AI) from pilot projects into daily operations, according to digital infrastructure company Equinix.
In an email interview with Newsbytes.PH, Equinix Philippines managing director Ireen Go-Catane said adding computing capacity alone would not be enough to support wider enterprise AI adoption.
“The most significant bottlenecks are the availability and predictability of power, grid-connection timelines, and high-capacity domestic connectivity,” Go-Catane said.
AI facilities also require infrastructure that can accommodate higher power consumption and more demanding cooling requirements, particularly for deployments using graphics processing units (GPUs).
Go-Catane said the country’s AI infrastructure needs span reliable electricity, subsea cables, domestic fiber networks, cloud connections, and data centers that allow businesses to connect with different telecommunications and technology providers.
Subsea cables link Philippine enterprises to overseas cloud and AI resources, while domestic fiber networks extend access across the archipelago.
Over the next three to five years, she said, priorities should include expanding reliable and sustainable power, improving domestic and international connectivity, and establishing predictable conditions for long-term infrastructure investment.
Private companies also need to invest in networks, renewable energy, cooling systems, and workforce capabilities, she added.
As companies integrate AI into fraud detection, automated decision-making, and customer service, the connections between their data, cloud platforms, and computing infrastructure become more consequential.
“If connectivity cannot keep pace, the issue is not simply slower performance. Enterprises may face higher data-transfer costs, inconsistent application experiences and greater difficulty scaling AI beyond isolated proofs of concept,” Go-Catane said.
“Fragmented data and manually managed connections can limit AI deployment at scale, regardless of available computing capacity.”
She said private connections between enterprise systems and technology providers could help businesses manage these problems by giving them more predictable performance and greater control over how their data travels.
Balancing local and overseas AI resources
Go-Catane advocated a distributed approach in which companies keep certain data and applications in the Philippines while accessing specialized computing resources elsewhere in the region.
“The Philippines benefits most from a distributed AI strategy that balances local inference with high-speed regional interconnection,” she said.
AI inference is the use of a trained model to generate responses, predictions, or other outputs. Applications requiring quick responses can benefit from having that processing closer to users and enterprise data.
Large model-training facilities, meanwhile, present near-term challenges involving grid capacity and liquid cooling, Go-Catane said.
Under a distributed setup, sensitive data could remain in an enterprise-controlled environment in the Philippines, while approved data or workloads connect to overseas AI infrastructure.
She said this could reduce the need to move entire datasets offshore and allow businesses to choose among cloud platforms, specialized AI providers, and other computing services.
Equinix uses its Manila data centers as local connection points to its international infrastructure. According to the company, its global network comprises more than 280 data centers across 77 metropolitan areas.
Its Equinix Fabric service provides software-defined private connections to cloud platforms, networks, and AI providers across markets.
Go-Catane said these connections could reduce exposure to public Internet congestion and give enterprises more flexibility over where they deploy workloads.
Bringing inference closer to enterprise data
The company is also offering Equinix Inference Exchange, which combines Nvidia-accelerated infrastructure, Together AI’s platform supporting more than 200 open-source models, and Equinix’s data centers and private connectivity services.
The service targets problems that can emerge when businesses depend on centralized AI infrastructure, including latency, data-transfer costs, security concerns, and reliance on a single cloud or model provider.
“By bringing inference closer to users, applications and enterprise data, the service reduces the distance that queries and responses must travel,” Go-Catane said.
She said the approach could shorten the time before an AI application begins returning a response, while reducing unnecessary data transfers and associated costs.
Potential Philippine applications include fraud detection and credit-risk assessment in financial services, clinical information analysis in healthcare, recommendation engines in retail, and AI assistants in telecommunications and e-commerce.
Logistics and industrial companies could also use the infrastructure to analyze data from connected equipment, anticipate maintenance needs, and monitor supply chains.
For privacy and security, Go-Catane said enterprises could use private connections and process information closer to where it resides, reducing the volume of sensitive data transferred across borders.
She said the broader goal is to give companies access to the computing, models, and connectivity needed to run AI applications without having to build and manage every component themselves.


