Summer 2026

Intern – AI/HPC System with Advanced Memory System Research

Posted on 1/26/2026

SK hynix

SK hynix

5,001-10,000 employees

Global producer of DRAM and NAND.

Compensation Overview

$26 - $50/hr

+ Housing Allowance

San Jose, CA, USA

In Person

Category
AI & Machine Learning (1)
Required Skills
LLM
Python
Tensorflow
CUDA
Keras
Pytorch
C/C++
Requirements
  • Accepted or currently enrolled in a PhD/MS program in Computer Science or a related STEM field
  • Expected to complete degree by July 2027 or earlier for PhD’s and 2026 for MS candidates
  • Hands on experience with ML/DL frameworks like TensorFlow/Keras or PyTorch
  • Excellent programming skills in Python, CUDA, and C
Responsibilities
  • Demonstrate feasibility of AI system with advanced memory solutions (HBM, CXL memory)
  • Excellent in reading research papers and implementing them to code
  • Build and evaluate the advanced memory system for an LLM training/inferencing

Summary of SK hynix: 1) What does the company do? It is a global producer of semiconductor memory, specializing in DRAM and NAND Flash that power IT devices like computers, phones, and servers. 2) How do its products work? DRAM stores data using tiny capacitors and requires periodic refreshing to keep data, providing fast volatile memory. NAND Flash stores data in non-volatile memory cells and preserves data without power, serving as durable storage. Both chips are mounted in devices to enable fast processing and data retention. 3) How is the company different from competitors? It ranks as the second-largest memory chip maker with a broad product portfolio and strong technology, backed by the SK Group, giving it scale, resources, and global reach. 4) What is the company's goal? To become the world’s best semiconductor company by meeting diverse customer needs and leading in memory technology.

Company Size

5,001-10,000

Company Stage

IPO

Headquarters

Seoul, South Korea

Founded

1983

Simplify Jobs

Simplify's Take

What believers are saying

  • Multi-year Nvidia partnership expands revenue beyond DRAM into AI platforms.
  • HBM4 qualification for Nvidia broadens SK hynix's addressable next-generation supply.
  • South Korea's gigawatt-scale AI cloud builds local demand through 2027.

What critics are saying

  • Samsung and Micron HBM4 qualification dilutes SK hynix pricing power.
  • Nvidia can pressure SK hynix on capacity, yields, and delivery schedules.
  • Wafer shortages until 2030 constrain output if fab expansion slips.

What makes SK hynix unique

  • SK hynix is Nvidia's largest memory partner, central to AI infrastructure.
  • June 2026 partnership covers Vera Rubin, Vera CPUs, RTX Spark, and Jetson Thor.
  • SK hynix uses Nvidia Omniverse, CUDA-X, and PhysicsNeMo for fab automation.

Help us improve and share your feedback! Did you find this helpful?

Your Connections

People at SK hynix who can refer or advise you

Benefits

Health Insurance

Paid Vacation

Paid Holidays

Paid Parental Leave

401(k) Company Match

Educational reimbursement

Donation Matching and volunteering opportunities

Corporate discount programs

Health Insurance

Paid day offs: PTO + Company Holidays + Happy Fridays

Growth & Insights and Company News

Headcount

6 month growth

-2%

1 year growth

-2%

2 year growth

-2%
Asharq Business
Jun 8th, 2026
Nvidia CEO: sell-off in technology stocks is a buying opportunity

Nvidia CEO: sell-off in technology stocks is a buying opportunity Huang: AI sector is still in the early stages of building its infrastructure Reading time - 3 minutes Published: June 7, 2026 20:54 * South Korea's Kospi index fell as investors cut their bets on AI stocks * Nvidia has signed a multi-year agreement with SK Hynix to manufacture advanced generations of memory chips * Huang: You should be happy because you can buy stocks at discounted prices Jensen Huang, CEO of Nvidia, said the technology stock sell-off represents a buying opportunity, emphasizing that AI infrastructure development is in its early stages. Nvidia and SK Hynix announced an agreement to develop AI memory chips. Huang described AI as set to become a global infrastructure like the internet. Jensen Huang, CEO of Nvidia, said that the sell-off that has hit global technology stocks since last week represents a buying opportunity, stressing that AI infrastructure development has only just begun. South Korea's benchmark Kospi index fell on Monday as investors reduced their bets on AI stocks that had fueled the rally in global equity markets. Also, concerns about AI overvaluation pressured technology stocks globally, while their US counterparts fell on Friday due to fears of possible interest rate hikes. Responding to questions about how to view the recent downturn, Huang said the sector is still in the early stages of building the infrastructure that will form the foundation of the AI-driven future. Agreement between Nvidia and SK Hynix Nvidia and SK Hynix announced on Monday that they had reached a multi-year agreement to develop future generations of memory chips dedicated to AI, a win for the South Korean company competing with Samsung Electronics in the fast-growing sector. Shares including SK Hynix pared some losses after South Korean President Lee Jae-myung said on Monday that he believes the local market is still undervalued. Huang said: "We are at the beginning of this phase, and whatever happened in the stock market, you should be very happy because now you can buy at discounted prices." He added: "Everyone should be very excited," in comments to reporters after meeting with SK Group Chairman Chey Tae-won in Seoul. AI as global infrastructure Like many of his peers in the technology sector, Huang has repeatedly emphasized that AI will drastically transform wide sectors of the global economy and change how people work and live. This is expected to drive demand for data centers and chips needed to power future AI services to high levels. He said: "It is well accepted that AI will become infrastructure for the world, just as the internet became infrastructure for the world."

Mousechief
May 5th, 2026
Manufacturers have started developing DDR6 RAM with an eye on 2028.

Manufacturers have started developing DDR6 RAM with an eye on 2028. By Daniel Whitaker Game Journalist at 7GrandSteps | Reviews and Gameplay Articles Last updated: May 05, 2026 Samsung, SK hynix, and Micron - the biggest names in memory - have quietly started work on DDR6 standardization. JEDEC hasn't signed off on final specs yet, but companies are already coordinating with substrate suppliers to get test samples out the door. The initial clocking target is 8,800 MT/s (roughly on par with top DDR5 modules), with a roadmap that hopes to push toward 17,600 MT/s later on. To keep signals usable at those rates, the proposed physical approach shifts from DDR5's 2x32-bit lanes to a 4x24-bit arrangement, i.e., narrower lanes but more parallel channels. Classic DIMM connectors create real physical limits, so the supply chain is looking to CAMM2 as a form factor that preserves signal paths better - whether it becomes dominant depends on board-level adoption and vendor buy-in. Expect early customer validations in 2027; broader commercialization and mass shipments are penciled in for 2028-2029, primarily aimed at AI DCs (AI data centers). That timeline still assumes a lot of integration work (PHY IP, thermal behavior, supply logistics, etc.), so consider these dates provisional rather than carved in stone.

이데일리
Apr 14th, 2026
SK hynix's Solidigm completes $100M US R&D investment 18 months ahead of schedule amid AI-driven NAND demand surge

SK Hynix is accelerating investment in its US NAND flash subsidiary Solidigm as AI-driven demand for storage grows. The company has deployed over $100 million to its Sacramento research and development centre, completing its five-year investment target 18 months ahead of schedule. Solidigm, acquired from Intel in 2020 for approximately $9 billion, posted its first annual profit in 2024 after three years of losses. The turnaround comes as AI datacentres drive surging demand for enterprise solid-state drives, Solidigm's main product. NAND prices have risen for 15 consecutive months, reaching record highs. SK Hynix has climbed to second place in the SSD market through Solidigm. The company recently opened an office in San Jose to strengthen collaboration with major technology clients including Nvidia, Apple and Google.

AInvest
Apr 11th, 2026
SK hynix invests in memory-centric AI infrastructure amid US patent infringement probe

SK hynix has invested in Semidynamics to co-develop memory-centric AI infrastructure optimised for data movement and access efficiency in large language models and AI inference applications. The partnership aims to align memory technologies with AI processing demands to improve performance and scalability. However, the company faces a US International Trade Commission investigation launched on 26 March 2026 into alleged patent infringement involving certain NAND and DRAM memory chips. The probe, following a complaint by MonolithIC 3D Inc., could affect SK hynix's access to the US market. Investors are evaluating the long-term benefits of the AI infrastructure partnership against near-term regulatory and intellectual property uncertainties. The ITC has not yet determined the investigation's merits.

Tech in Asia
Apr 8th, 2026
SK hynix to supply Dell with 2TB QLC AI drives from April

SK hynix has developed new high-capacity AI computer drives and will begin supplying Dell Technologies in April as its first global IT enterprise partner. The memory chipmaker is offering 1TB and 2TB versions using 321-layer QLC NAND technology, which stores four bits per cell to increase capacity. The technology delivers up to 56% faster writes and over 23% better write power efficiency compared with previous generations. IDC expects QLC NAND to comprise 61% of the global PC and notebook SSD market by 2027, up from 22% last year. SK hynix plans to extend 321-layer NAND to enterprise SSDs for data centres and Universal Flash Storage for smartphones. The company is pursuing petabyte-scale SSDs that could replace hard disk drives for storing large AI datasets.

INACTIVE