By early 2016, HTC, Xiaomi, and Lenovo had all quietly walked away from the same chip. MediaTek’s Helio X20 was supposed to be the flagship that finally put the company in the same conversation as Qualcomm, the world’s first deca-core phone processor, ten cores in three clusters, a genuinely clever idea. It also ran hot on TSMC’s 20nm node, the exact process that had just torched the reputation of Qualcomm’s Snapdragon 810, and the big customers bailed before launch. That is the Helio era in one image, and I find it a little painful because the ideas were usually good and the execution kept letting them down.
The deca-core thing was classic MediaTek: a spec you could put on a box that no rival could match, sold to a market that counts cores. Two Cortex-A72 up top, then two islands of A53, all fed through a tri-cluster scheduler meant to fix a real inefficiency in the usual big.LITTLE split. On paper, it was ahead of its time, and here is the twist that still amuses me: the industry eventually agreed. The one-big, three-middle, four-little layout every Dimensity flagship uses today is the same three-tier thinking MediaTek got mocked for in 2015. The problem back then was the GPU, a Mali that trailed Adreno badly, plus a thermal story that meant the chip benchmarked like a flagship and then throttled like a mid-ranger the moment a game ran long. The X25 was just an overclocked X20 that Meizu took as a semi-exclusive. The X30 moved to 10nm and was genuinely better, and it shipped in essentially one phone, the Meizu Pro 7. After that, MediaTek stopped making 10-core CPUs and, more importantly, stopped chasing the flagship at all. Qualcomm kept the top of the market by default.
The retreat that actually worked
Giving up on the crown turned out to be the smartest thing MediaTek did in this stretch. It went back to what it has always been good at, the mid-range, and quietly made that tier more interesting. The Helio P60 in early 2018 is the one that mattered. Built on 12nm, it was the first MediaTek chip with a dedicated multi-core APU and the NeuroPilot software to feed it, which is to say MediaTek put real on-device AI into affordable phones a full generation before most people were talking about NPUs. It sold enormously in India and Southeast Asia, the markets MediaTek never actually lost. The P-series was never going to win a benchmark shootout, and it did not need to.
The move that defined the back half of the decade was rebranding. In July 2019, MediaTek took its mid-range silicon, bolted on a suite of latency and thermal tricks called HyperEngine, and called the result a gaming chip: the Helio G90 and G90T. It was mostly marketing, an A76 and Mali-G76 setup on the same 12nm process, but it was smart marketing, because the phone that carried it was Xiaomi’s Redmi Note 8 Pro, the first time Xiaomi trusted a top-end MediaTek part in a phone people actually wanted. I put the G90T straight up against Qualcomm’s Snapdragon 730G back then, and the interesting result was that it was a real fight, not a value-brand apology. From there, the gaming badge spread across the whole stack. MediaTek pushed it down to entry level with the G35 and G25, climbed back up with the G95, and kept the line fresh into the next couple of years with the G96 and G88. None of these were flagships. All of them sold, and they ran as the 4G gaming workhorses right alongside the new 5G Dimensity push that was starting at the same time.
What the Helio years taught MediaTek, and what it seems to have to relearn every few years, is that it is not a flagship company; it is a volume company, and the two require completely different temperaments. The deca-core theater of 2015 to 2017 was the company forgetting that. The gaming pivot of 2019 was its remembering, right before 5G handed it a reason to climb the mountain all over again. I’m skipping the benchmark-integrity mess that trailed some of these chips, which deserves its own uncomfortable post. What I keep thinking is that MediaTek spent this whole era being told it could not build a flagship, made peace with that, got rich on the mid-range anyway, and then immediately went and built a flagship. That story is the next chapter.
Two decades of MediaTek, oldest to newest: Turnkey Empire (2006-2015) · Helio Years (2015-2019) · Dimensity 5G (2019-2021) · Flagship Assault (2022-2023) · The AI Turn (2024-2026)